Embed
Email

Comparing the General Education Development _GED_ Tests to the ...

Document Sample

Shared by: yaohongmei
Categories
Tags
Stats
views:
1
posted:
1/10/2012
language:
pages:
122
Comparing the General Education Development (GED) Tests to the



ACT Computer-Adaptive Placement Assessment and Support System (COMPASS)



Placement Tests



As Predictors for College Readiness



CHAPTER ONE



INTRODUCTION



Postsecondary institutions are set squarely in the crosshairs of one of the most



challenging issues facing this country’s employers for at least the next generation. That



challenge is to educate and train enough qualified and competent workers to do the jobs



necessary for the nation to remain the world’s economic leader. Without enough students



attending college, the country can not hope to graduate enough engineers, teachers,



business professionals, computer programmers, technicians, and other occupations



requiring college level training to fill the demand that employers will experience in the



future (Spangenberg, 2005). To satisfy that need, postsecondary institutions more than



ever must be prepared to enroll and graduate students from nontraditional groups who are



often unprepared for college-level work (Lamkin, 2004).



Along with other groups of nontraditional students, completers of the General



Education Development (GED) examinations increasingly access higher education



through community colleges (Baycich, 2003). Like other nontraditional students, GED



completers often do not have the academic skills necessary to successfully persist and



complete either a degree or occupational certificate (CALEC, 1993b; NCWE, 2004;



Pusser, 2007; Zafft, 2006). Enrollment in fee-based developmental or remedial courses



designed to improve students’ academic proficiency extends the time that underprepared

2

GED completers must persist in college. These courses do not result in college credit



toward a degree or occupational certificate and consume students’ often limited financial



resources. These dual pressures decrease the likelihood that nontraditional students will



successfully complete a college degree or certificate (Hoyt, 1999).



Closing the Gaps



Business, political, and educational leaders in Texas recognize the urgency for



colleges and universities to recruit, retain, and graduate more students. “Closing the



Gaps” is the master plan developed by the Texas Higher Education Coordinating Board



(THECB) to increase enrollment in Texas colleges and universities to a level



commensurate with other large population states (Arnone, 2003b). Currently, the



percentage of Texans who attend college is considerably lower than that of other states



with large populations. In 2003, only 4.9 percent of Texans attended college as compared



to states like California with 6.1, Illinois with 6.0 and 5.6 in New York. These large states



represent Texas’ main competition for economic development and growth. Currently, the



percentage of Texans who attend college, while improved, is still considerably lower than



that of other states with large populations. Comparing 2006-2007 total enrollment figures



to total state population estimates for 2006, college attendance rates for those states are:



California, 6.67%; Illinois, 6.47%; New York, 6.02% and Texas 5.28% (IES, 2007; US



Census, 2008). Among postsecondary education leaders the worry remains that the



Texas college attendance rate is predicted to decline to less than 4.6 percent by 2015



unless interventions and initiatives of the THECB and legislature are successful (Arnone,



2003b).

3

Begun in 2001, the “Closing the Gaps” plan calls for 500,000 additional students



to be enrolled in Texas’ colleges and universities by the year 2015 as compared to the



benchmark year of 2000 (Arnone, 2003a). According to demographic projections, college



enrollment over that same period, 2001-2015, is expected to increase by at least 200,000



students without any special effort by the state’s higher education community. To meet



the state’s goals for enrollment and graduation, most of the remaining 300,000 students



likely will have to come from the ranks of nontraditional students like GED completers



and the state’s growing Hispanic population. (Arnone, 2003b; CALEC, 1993b).



GED Completers and Closing the Gaps



While more than 40,000 Texas dropouts complete the GED each year, some



surveys show that up to 66% of them cite attending higher education as a personal goal



(Baycich, 2003; NCWE, 2004). For this reason, the GED population is clearly an



important one for Texas to consider as it attempts to meets its “Closing the Gaps” goals



(Baycich, 2003; CALEC, 1993b, 1995; Soltz, 1996). Further, a mandate by the Texas



Legislature in 2007 that the Texas Adult Education System be aligned with higher



education clearly indicates that the state recognizes the importance of transitioning GED



completers into postsecondary education (CALEC, 1993a; LBB, 2007).



The Texas Success Initiative



The Texas Success Initiative (TSI) was adopted by the Texas Higher Education



Coordinating Board (THECB) in December of 2003 with the purpose of improving the



performance of individual college programs and to ensure the success of students in



higher education (CALEC, 1993a; TAC, 2003). The TSI establishes a system of student



assessment practices designed to predict college readiness and college success for all

4

students entering postsecondary education in Texas. Most importantly, the legislation



identifies several assessment instruments that are acceptable by the state for determining



college readiness for students. While certain exemptions apply, the TSI legislation sets



the minimum scores for the domains of the respective instruments that higher education



institutions must adhere to when determining students to be college ready. Institutions



however may demand higher scores locally if they so choose. Among the tests approved



as acceptable for TSI purposes are the ASSET and COMPASS offered by ACT, the



ACCUPLACER offered by the College Board, and the Texas Higher Education



Assessment (THEA) formerly known as the Test of Academic Skill Proficiency (TASP)



(TAC, 2003). Each of the tests assesses the skills of students in three domains: reading,



mathematics, and writing. Students scoring at or above the established benchmarks for



each respective domain are considered college ready and may be placed in coursework



that apply toward the completion of an associate’s or baccalaureate degree or an



occupational certificate. For instance, a student scoring 81 out of possible score of 99 on



the COMPASS Reading Placement Test is considered as being college ready. To meet



college readiness standards for mathematics, a student must score at least 71 out of a



possible score of 99 on the COMPASS Algebra Test. Students who score below the



established benchmarks for a particular domain must be placed in remedial or



developmental courses to address the students’ respective skill deficiencies. Remedial or



developmental courses are non-degree credit courses that do not apply toward the



completion of any certificate or degree (TAC, 2003).

5

GED Completers and the Texas Success Initiative



The purpose of developmental courses is to remediate underprepared students so



they can be admitted into a course of study that will terminate in an associate’s degree,



occupational certificate or transfer to a four-year institution. However, the data continue



to reveal that students who enroll in developmental courses and matriculate on to



complete a course of study at a community college or transfer to a four-year university,



do so at single-digit rates (Pathey-Chavez, 2005).



Because students preparing for the GED possess many of the characteristics



identified as being barriers to success in college, postsecondary institutions and especially



community colleges, need to create early identification processes that will help GED



completers overcome those barriers. A method of predicting college readiness of GED



completers based on their GED Test scores could help align the curricula of GED



preparation courses with community college developmental and college level courses.



Students preparing for the GED who wish to enter college could be advised to remain in



no cost GED preparation classes available through the Texas Adult Education System



and attempt the GED examinations only after completing the curricula more likely to



prepare them to be college ready. By remaining in no cost adult education classes and



achieving college readiness before completing the GED and entering college, those



students would save a considerable amount of money by avoiding costly developmental



coursework (HCCS, 2004; Hoyt, 1999; Roth, Crans, & Carter, 2000).



The Texas Adult Education System



Traditionally, community colleges act as a gateway to postsecondary education



for many students who drop out of high school before graduation. To their advantage,

6

many community colleges in Texas are recipients of federal and state grant funds that



enable them to provide free GED preparation classes (TCALL, 2007). While separately



acting as certified GED Testing centers as well (TEA, 2007a, 2007b), Texas community



colleges are in particularly advantageous positions to use the Texas Adult Education



System as a natural pipeline for dropouts to access higher education.



Statement of the Problem



Even though providers in the Texas Adult Education System, which includes



many community colleges, have as one of their performance measures, the transition of



GED completers into postsecondary training or education (TEA, 2007a), a review of the



literature reveals that there is no research-based information available suggesting that the



current version of the GED examination could be used to predict college readiness



(Hamilton, 1998; Rose, 1999; Smith & Goetz, 1988; Wolf, 1983).



Purpose of the Study



The purpose of this study is to determine if scores from the GED Tests can be



used to predict college readiness. First, the study compares the content, constructs, and



reliability of the COMPASS Reading and Mathematics Placement Tests to the GED



Reading and Mathematics Tests respectively. If the tests are found to be sufficiently



similar, it will then determine if the scores of the two tests can be meaningfully linked.



Using a correlation and linear regression analysis, the study will determine the strength of



the relationships between the scores of the two tests with GED Test scores acting as the



independent variables and COMPASS Test scores acting as the dependent variables.



Finally, if the scores of the two tests possess linkages of sufficient strength, a



concordance of the scores will be produced for the overall study group and by subgroup

7

based on gender, ethnicity, and gender and ethnicity in combination. If the study confirms



that meaningful linkages exist between the scores of the two tests, it could provide much



needed researched-based data for postsecondary institutions as well as adult education



researchers and practitioners regarding the prediction of college readiness for GED



completers. Accurate prediction of college readiness for GED completers based on their



GED scores could result in better alignment of postsecondary requirements for college



readiness and the curricula used to prepare adults learners to take the GED.



Significance of the Study



GED completers are an important constituency group for community colleges but



are among those populations that are likely to be underprepared for success in college



(Soltz, 1996). Early identification of GED completers who are likely to struggle to be



successful in higher education would allow community colleges to focus resources and



supportive services and help them to successfully complete a course of study. If the GED



Tests and COMPASS Tests are measuring similar content, similar constructs, and are



similarly reliable, it increases the degree to which their scores can be linked. If GED Test



scores can be linked to COMPASS Test scores, it suggests that they can be useful for



predicting college readiness for GED completers. Accurate prediction of college



readiness using GED Test scores will benefit students by informing them if they possess



the academic skill proficiencies required to enter college level work. If they lack college



readiness skills they can then be advised accordingly.



This study is significant for three reasons. First, it supports an initiative by the



Texas Higher Education Coordinating Board (THECB) to align Adult Education



instruction with higher education and increase the enrollment and graduation rates of the

8

states’ colleges and universities. Second, the study informs Adult Education researchers



and practitioners regarding the relationship of the content and reliability of the GED



Reading and Mathematics Tests relative to the content, constructs, and reliability of the



COMPASS Reading and Mathematics Placement Tests. If the tests are similar in these



three areas, it suggests that their scores may be meaningfully linked. Finally and most



importantly, if the study confirms meaningful linkages exist between the scores of the



GED Reading and Mathematics Tests and the scores of the COMPASS Reading and



Mathematics Placement Tests, such linkages have the potential to accurately predict



college readiness for GED completers and would be of great value to institutions of



higher education by helping them to align GED preparation curricula with college



readiness standards.



Research Questions



For GED Reading and Mathematics Test scores to be reliably predictive of scores



on the COMPASS Reading and Mathematics Placement Tests, the respective tests must



exhibit a substantative degree of similarity. Dorans and Holland (2000) hold that there are



five important test criteria to consider when assessing test linkability and test score



linkage. Those criteria are (1) tests and their scores should not be equated if they measure



different constructs; (Dorans & Holland, 2000); (2) tests that measure the same



constructs should not be equated if they differ in reliability; (3) the functions that equate a



first test to a second, should work equally well in reverse when equating scores from the



second test to the first; (4) if two tests can be sufficiently equated, it should not matter to



the individual which test they take; and (5) the degree to which two tests are equated



should not vary by subpopulation (Dorans & Holland, 2000). For their purposes, Dorans

9

and Holland refer to “linking” as the function(s) that can be used to connect the scores of



one test to the scores of a completely different test. They refer to “equating” on the other



hand, as meeting the five criteria described earlier when applied to different versions of



the same test. They also suggest that a common method used to determine the constructs



measured by a test consists of inspection of the content of the tests and how the tests’



items are worded (Dorans & Holland, 2000). Given the lack of systematic theory in this



field, it is important that linking studies be conducted for individual tests pairs like the



GED and COMPASS to determine the degree to which they can described as linkable and



thus provide a sense of their usefulness for predictive purposes (Dorans & Holland,



2000).



In considering the technical relationship between the GED Reading and



Mathematics Tests and the COMPASS Reading and Mathematics Placement tests, the



following research questions are asked:



1. To what degree are the tests measuring the same content?



2. To what degree are the tests similarly reliable?



3. To what degree are the tests symmetrical?



4. To what degree are the tests’ scores linkable?



5. To what degree are any test score linkages population invariant?



If the GED and COMPASS Tests exhibit enough similarity in content and reliability then



further research into the degree to which their scores are linkable could yield meaningful



results. If the tests are too dissimilar, then a linking study of their scores is less likely to



reveal meaningful relationships (Kolen & Brennan, 2004).

10

Hypotheses



The hypotheses posited for the technical relationships between the GED Reading



and Mathematics Tests as compared to the COMPASS Reading and Mathematics



Placement Tests are as follows.



1. The content measured by the GED Reading and Mathematics Tests will be



similar to those of the COMPASS Reading and Mathematics Placement Tests



to the degree that their scores can be meaningfully linked.



2. The reliability scores for the GED Reading and Mathematics Tests will be



similar to those of the COMPASS Reading and Mathematics Placement Tests



to the degree that their scores can be meaningfully linked.



3. Linkages between GED Reading and Mathematics Test scores and their



respective COMPASS Reading and Mathematics Placement Tests scores will



be symmetrical to a meaningful degree.



4. GED Reading and Mathematics Tests scores will be linkable to their respective



COMPASS Reading and Mathematics Placement Test scores to a meaningful



degree.



5. GED Reading and Mathematics Test score and COMPASS Reading and



Mathematics Placement Test score linkages will be population invariant.



This study consists of a content analysis of the GED and COMPASS Reading and



Mathematics Placement Tests along with a correlation and linear regression analysis of



the tests’ scores. If the scores of the two tests can be sufficiently linked, an equipercentile



scaling of those scores will be used to construct a series of concordance tables that will

11

describe the relationship of those scores overall and by subgroup based on gender,



ethnicity, and a combination of gender with ethnicity.



Definition of Terms



For the purpose of this study the following terms are defined.



Adult Education – federal and state grant funded programs that provide basic



literacy skills, English as Second Language (ESL) preparation to out-of-



school adults (TEA, 2007b).



Adult Basic Education – basic skills training for out-of-school adults that is



equivalent to a grade K-8 level of proficiency, including ESL (TEA,



2007b).



Adult Secondary Education - basic skills training for out-of-school adults that is



equivalent to a grade 9-12 level of proficiency (TEA, 2007b).



“Closing the Gaps” - the master plan by the Texas Higher Education



Coordinating Board (THECB) designed to increase college enrollment in



Texas to a level commensurate with that of other large population states



(TEA, 2007b; THECB, 2000).



College Readiness – ability of a student to successfully produce college level



work (TAC, 2003).



College Success - completion of a postsecondary certificate or degree (TAC,



2003).



Developmental Studies – series of college courses designed to remediate



academically underprepared students to successfully produce college level

12

work. Developmental studies are fee-based courses but developmental



studies credits are not transferable to a college degree (TAC, 2003).



Placement – system of assessing college students’ academic proficiencies for



appropriate enrollment in developmental or standard college credit course



(HCCS, 2004).



Organization of the Study



Having discussed the significance of the study, its research questions, and



hypotheses, the remainder of the study will include a review of the literature pertinent to



this topic, a methodology section describing the study’s analytical techniques, a results



section that presents the study’s findings, and a final section that will describe



conclusions and implication for changes to current practice including advisement,



curricula, and college readiness determination indicators for GED completers that might



be derived from those findings. The literature review will describe research that relates to



accountability and performance for postsecondary institutions, in particular community



colleges and their relevance to this topic. It will also include a brief discussion of the test



equating and score linking practices along with a description of the development, history



and characteristics of the COMPASS and GED Tests. The methodology section will



provide an overview of the processes for conducting the content analysis of the



applicable COMPASS and GED Tests and describe the quantitative techniques used to



determine the relationships between the tests’ scores. The results section will discuss the



outcome of the content analysis procedures and the subsequent analysis of test score data.



The study’s final section will layout conclusions that might be construed from the



findings and their implications for students, researchers, policymakers and practitioners.

13

Implications for changes to current practice might in include (1) changes to how GED



completers are advised when applying for college admission, (2) changes to curricula that



prepare adult learners for the GED examinations, and (3) changes to the college-readiness



indicators accepted by Houston Community College for GED completers.

CHAPTER TWO



REVIEW OF THE LITERATURE



The literature review for this study is divided into three sections. The first section



will discuss findings regarding postsecondary institution performance accountability and



describe its relationship to college readiness and college success for students. The second



section of the literature review will be focused on an overview of test equatability and



test score linkage considerations, methods, and techniques. The final section of the



literature review will discuss material specific to the COMPASS and GED Tests



regarding their history, content, constructs measured, reliability, and validity.



Postsecondary Institution Accountability



Legislators clearly recognize the role of community colleges and other



postsecondary institutions in maintaining the health and vitality of the nation’s economy.



At the same time, decision makers also demand that those institutions continually



demonstrate that they are worthy of the taxpayer’s largess. An important measure of



institutional effectiveness is the number of students who complete their course of study



within a prescribed time period and graduate. Postsecondary institutions, whether private



or public, can find their funding in jeopardy if they do not adequately retain and graduate



their students (HCCS, 2004; Lau, 2003). The Spelling’s Commission on the Future of



Higher Education describes postsecondary education as costing too much while



producing too little (ASSCB, 2006; Carey, 2007). Coupled with a national movement by



state governments to implement various outcome-based funding mechanisms,



accountability and performance are words that are becoming more important in the



vocabularies of community college administrators and faculty. Accordingly,

15

administrators and faculty in higher education in particular are becoming more sensitized



to implications of systems that measure their performance (Burke & Minassians, 2004;



McLendon, Hearn, & Deaton, 2006).



While not yet as rigorous and demanding as the performance measures placed on



public schools, the expectation that community colleges function efficiently and measure



student outcomes instead of resource inputs is a reality (Burke, 2005). Measurement of



resource inputs refers to for example; the number of hours a particular retention program



was provided to students or how many students participated in the program. These kinds



of descriptive measures are coming under increasing criticism however, because they do



not inform community college decision makers, legislators, parents, and students of the



degree to which programs were effective at improving student achievement, retention,



and completion (Burke, 2005).



While there is little agreement among state legislatures regarding what specific



accountability measures should be imposed on community colleges, there is a clear



consensus that the time for accountability has arrived (Burke & Minassians, 2004). A



survey of states’ accountability systems conducted by Burke and Minassians (2004)



found that there was little consistency among those states’ performance reporting



indicators. Additionally, the survey further revealed that the indicators that do exist



generally continue to stress inputs more than outcomes. As a result, accountability



systems currently have little impact on policymaking at the campus level and large



majorities of the leaders in college divisions, and departments have little or no familiarity



with how their respective institutions define its accountability system (Burke &



Minassians, 2004).

16

Predicting College Readiness and Success



The literature shows college success is defined by postsecondary institutions in a



variety of ways. Not surprisingly, the methods used by these same institutions to predict



student success range widely as well. Conley (2007) in a report for the Bill and Melinda



Gates Foundation describes the complications involved in determining college readiness



for students. The report discusses use of traditional student attributes such as grade point



average, course titles in transcript, and standardized test scores but concludes that a much



more robust system of indicators is needed for students to know where they stand in



regard to their academic preparation for college (Conley, 2007).The literature generally



identifies college success for students three ways: academic achievement, retention, and



completion (Conley, 2007; Lau, 2003; Schmid & Abell, 2003; Wolf, 1983; Wyman,



1997). Academic achievement is regularly measured using Grade Point Average (GPA)



(Carlan & Byxbe, 2000; Conley, 2007; Ridgell & Lounsbury, 2004; Spitzer, 2000;



Stovall, 2000; Zamani, 2000). Retention refers to keeping students actively enrolled in



college within a given semester and as would be intuitively expected, there a consistent



relationship between academic achievement and retention (DeBerard, Spelmans, & Julka,



2004). Retention correspondingly is closely connected to completion. Completion refers



to the attainment of some terminal credential, like a certificate or diploma (Lau, 2003).



These three measures, retention, completion, and academic achievement and how they



are defined, are all major factors for decision makers to consider for inclusion in any



accountability system established for institutions of higher education (ASSCB, 2006).

17

College Success and Personal Characteristics



Hoyt (1999) studied several student personal characteristics that might be useful



as predictors of academic performance and future college success. This study supported



previous findings that determined poor academic performance during the first semester of



college could be attributed to minority status, and working fulltime. The study suggested



that living at home while attending a community college appeared to have a significant



positive effect on student retention and course completion, while there appeared to be a



significant negative relationship between the level of remediation required by students



and retention (Hoyt, 1999).



Miglietti and Strange (1998) studied how student performance related to age,



student learning style, and teaching style preference. The results of this study showed age



appears to have no significant effect on classroom environment or learning style



preference, although adult students expressed a significant preference for learner-centered



teaching styles in mathematics. Learner-centered instruction was found to be clearly



related to higher grades, a greater sense of accomplishment as well as a greater overall



satisfaction by both adult and traditional students (Miglietti & Strange, 1998).



Ridgell and Lounsbury (2004) looked at the relationship between college success,



general intelligence, work drive, and five personality traits. The study defined college



success by the course grade received by students in an introductory psychology class and



the students’ self-reported college GPA. General intelligence was defined using an



instrument developed by the two researchers which produced scores that were correlated



to the Otis-Lennon Test of Mental Maturity. The five personality traits-Extraversion,



Emotional Stability, Agreeableness, Conscientiousness, and Openness to Experience-

18

were measured using the Personal Style Inventory (PSI), another instrument developed



by the researchers. Work Drive similarly was measured with the researchers’ own



instrument. The results of a hierarchical multiple regression analysis indicated that



General Intelligence and Work Drive both exhibited a significant positive correlation



with college GPA and course grade. Only one personal characteristic, Emotional Stability



was related to course grade (Ridgell & Lounsbury, 2004).



Spitzer (2000) compared predictors of college success for traditional (age 23 and



under) and nontraditional (age 25 and over) fulltime undergraduate students. Five



personal characteristics including academic self-efficacy, global self-worth, social



acceptance, career decision-making self-efficacy and social support as well as two



learning characteristics intrinsic motivation and self-regulation along with two college



success measures (GPA and career decidedness) were compared using a multiple



regression analysis. Self-efficacy or belief in one’s own ability showed to be the strongest



positive predictor of GPA. Self-regulation and social support also had strong positive



predictive qualities relative to GPA. GPA was also positively correlated for females with



a strong self-regulating trait. Traditional students with high global self-worth and social



acceptance showed a negative correlation for GPA. Career decidedness showed mixed



results relative to the variables in the study (Spitzer, 2000).



DeBerard, Spelmans, & Julka (2004) conducted research to determine the



relationship between the college success measures of academic achievement (GPA) and



retention to ten health related characteristics, previous academic record, and personal



characteristics. The study’s results showed a significant positive relationship between



cumulative GPA and retention. GPA was also shown to have substantial correlations with

19

the ten health-related variables used in the study. The ten health-related variables



accounted for 56% of the variance of first year cumulative GPA while SAT scores alone



were shown to account for only 25% of first year cumulative GPA variance (DeBerard,



Spelmans, & Julka, 2004).



A study by Dozier (2001) looked at predictors of college success for international



students. The study suggests that legally documented international students are better



prepared academically and perform better academically than the general student



population enrolled at community colleges. However, college success for international



students is complicated by their language proficiency characteristics. International



students who have entered the country legally are more likely to require language



acquisition assistance earlier in their college experience than their undocumented



counterparts because legal entrants must exhibit English proficiency before being



admitted to the country and into postsecondary education. This variation in language



ability influences considerably college success for both groups. The study further



indicates that undocumented international students are so different in their personal



characteristics from documented international students that they should be studied and



evaluated separately when colleges are determining how to best improve the levels of



performance (Dozier, 2001).



In an attempt to develop a simple tool to predict future college success, a study by



Osborne (1997) looked at the correlation of several student personal characteristics and



college success. The study showed students who identified strongly with academics were



likely to achieve college success and those students who do not identify with academics



were at higher risk for problems in college related to their coursework. A moderate

20

relationship appeared to exist between academic identification and grade point average.



No relationship was identified regarding the students’ level of academic identification



and their likelihood of future graduation (Osborne, 1997).



Rosenthal and Wilson (2003) exploring how student psychological experiences



might effect college readiness, studied how being exposed to violence impacted the



academic performance of minority students who grew up in urban settings. The results of



this study suggest that a students’ exposure to community violence and academic



performance were not related. It did find that exposure to community violence was



related to an increased level of psychological distress for students during the first



semester of college and that increased levels of psychological distress were negatively



related to student persistence. Curiously, the study found no relationship between



psychological distress and grade point average (Rosenthal & Wilson, 2003)



Garavalia and Gredler (2002) conducted a regression analysis of how well four



self-regulated learning strategies along with student reliance on external sources for



learning guidance, cumulative grade point average, and aptitude predicted college success



for 256 psychology students at a southeastern college. The four learning strategies were:



(1) General Organization and Planning, (2) Environment Restructuring, (3) Recall



Ability; and (4) Typical Study Strategies. College success was defined by course grade



achievement. Results of the study indicated the four self-regulatory variables contributed



to 45% of course achievement (Garavalia & Gredler, 2002).



College Readiness and Academic Achievement



Correlating student performance on high school exit examinations and high



school course completion is a promising source of information that can be used to predict

21

college readiness. Determining what scores on applicable high school exit examinations



and which courses when completed by high school graduates will indicate college



readiness would be of great advantage to schools, colleges, students, parents, and



legislators. Such information would assist students, parents, and guidance counselors to



select high school courses that would be more likely to result in greater academic



preparedness by students (Roth, Crans, & Carter, 2000).



Roth, Crans, and Carter (2002) conducted a study that revealed completion of



high level mathematics and English high school courses were much more likely to predict



college readiness than high exit examination scores or student grade point averages.



While there was some concern by the researchers regarding a disparity in performance of



minority students relative to White students in English, overall it was clear that



completion of a high level mathematics or English course was a positive predictor of



college readiness. This prediction held true in the study even when students had received



poor grades in those high level high school courses. This work suggests that an important



way to improve on the number of college ready students is for schools and colleges to



place less emphasis on grade point average and exit test scores and more emphasis on



encouraging student completion of rigorous high school courses (Roth, Crans, & Carter,



2000).



Mulvenon, Stegman, and Thorn (1999) describe a study of 170 scholarship



recipients conducted for the purpose of developing selection criteria for a corporate



scholarship. The corporation that supported the study wanted to improve the likelihood



that the recipients of its scholarship would be successful in college. The literature review



associated with this study indicated that both cognitive and non-cognitive characteristics

22

have an effect on college success of students. Using previous scholarship recipients as a



study population, the researchers compared the college freshman year grade point



average to a modified high school grade point average (HSGPA) and ACT scores. The



modified HSGPA included academic courses only instead of all of the high school



courses completed by the students. College success for this study was defined as



achievement of a 3.00 GPA for the students’ college freshman year. While both ACT



score and class rank were identified as important predictors of college success, a stepwise



multiple regression procedure revealed that the modified HSGPA was a significant factor



in predicting freshman grade point average (FGPA). Additionally, the correlation



between the modified HSGPA and FGPA became stronger as HSGPA increased



(Mulvenon, Stegman, & Thorn, 1999).



Naumann, Bandalos, and Gutkin (2003) conducted research on how self-regulated



learning was predictive of college success for first-generation college students. The study



compared how well self-regulated learning variables and ACT Test scores predicted



college success for first and second generation college students. The study sample was



made up of 155 first and second generation college students at a large Midwestern



university. The results showed that for both first and second generation college students,



self-regulated learning variables accounted for more of students’ grade point average than



did ACT Test scores (Naumann, Bandalos, & Gutkin, 2003).



Popham (2006) discusses the reliability of using standardized test scores like the



SAT and ACT to predict student performance based on grade point average by students



once they have entered college. The author discusses claims that correlations between



these test scores and students’ success in college generally explains only about 25% of a

23

student’s grade performance. The other 75% of grade performance is purported as being



determined by other factors like motivation and study habits. The author concludes by



stating that colleges and universities should be more cautious regarding the use of SAT



and ACT for college entrance decisions given that other factors have three times more



influence on a student’s predicted performance (Popham, 2006).



Ting (1998) studied a small sample of 18 men and 36 women from low-income



families at a Midwestern public university to determine the correlation between ACT



score, class rank, eight psychosocial variables, and freshman year grade point average.



High school class rank, the psychosocial attributes of successful leadership experience,



and demonstrated community involvement were found to be the most effective predictors



of freshman year GPA for the group involved in the study (Ting, 1998).



Wolfe and Johnson (1995) studied 201 psychology students at the State



University of New York to determine the correlation between their SAT scores, high



school grade point average, 32 personality variables, and college GPA. The results of that



study showed that high school GPA accounted for 19% of the variance for college GPA,



while the personality trait of self-control and SAT score accounted for only 9% and 5%



of the variance respectively (Wolfe & Johnson, 1995).



Generally, the literature suggests that standardized assessments like the SAT and



ACT account for only a small portion of the variance for college readiness and college



success measurements. Personal characteristics like work ethic and self-efficacy along



with the type of academic preparation a student receives appear to be consistently more



influential on college readiness and success than are standardized test scores.

24

College Readiness and GED Completers



Individuals who complete the General Education Development (GED) test do so



for a variety of reasons. Although most report they had some employment, personal, or



family related goal that was met by completing the GED (Golden, 2003), an increasing



number of GED completers move on to postsecondary education according to the



American Council on Education. In 1967, only 36% of GED completers reported that



they planned to enroll into postsecondary education or training. That percentage soared to



49.7% in 1987. By the year 2000, nearly two-thirds of GED completers reported that they



planned to enroll into college (Baycich, 2003).



Postsecondary pursuits by GED completers range widely and include four-year



baccalaureate degrees, associate degrees, certificates, and short-term training programs.



According to the American Council on Education, which produces the GED, if a student



achieves an average score of 500 or higher out of a possible 800 on the five GED Tests



respectively, they would rank in the top 50% of high school graduates for that domain



(ACE, 2007b). College success by GED completers and how to predict it should then be



of great concern to colleges and universities. Community colleges especially should be



concerned because they enroll most of the GED completers who pursue postsecondary



education (Golden, 2003; Rose, 1999).



Fisher and Sandiford (2000) reference several studies conducted to examine both



the GED as a predictor of college success and to compare the performance in college of



GED completers to high school graduates. However, the studies cited all occurred well



prior to the latest version of the GED being put into place and as the GED evolves the



changes in its content could affect how well it might predict college performance. Using

25

a t-test, 146 matched pairs of high school graduates and GED completers based on



gender, race, and age range were compared by first semester grade point average and



total college grade point average. No significant findings were uncovered relative to the



performance of GED completers and high school graduates except in the area of



placement on probation. In this study, high school graduates were found to be



significantly more likely to be placed on probation than GED completers (Fisher &



Sandiford, 2000; O'Neill, 1995).



O’Neill (1995) likewise confirmed that there is no significant difference between



GED completers and Traditional High School (THS) completers relative to college



success and persistence. This study compared the Grade Point Average (GPA) of 47 GED



completers and 92 THS completers enrolled in an urban community college. The study



also emphasized the importance of GED completers to community colleges as a customer



base.



One source was identified as particularly pertinent to the subject of this literature



review. Tokpah and Padak (2003) compared the placement of GED completers and high



school graduates into remedial courses at Kent State University (KSU). The COMPASS



Test was used by KSU to determine appropriate placement of every student entering the



university, including GED completers. The results of the study indicated that both the



average GED completer and traditional entering college freshman were likely to place out



of developmental coursework in reading and English, but both groups were likely to be



placed in a remedial mathematics class. However, GED students were more than twice as



likely to be placed in a remedial mathematics classes than were traditional college



freshman (O'Neill, 1995; Tokpah & Padak, 2003).

26

Smith and Goetz (1988) studied using GED Test scores for placing GED



completers into classes at North Harris County College (now Lone Star College) in



Texas. This study looked at 1,344 GED completers who had taken the GED exams and



subsequently enrolled into semester credit courses at the college between the years of



1973 and 1985. GED Writing and ACT English subtest scores were correlated with



freshman composition course grades to determine if GED subtest scores could better



predict college success than the selected ACT subtest scores. During the period of time



that the study was conducted, North Harris County College used the ACT subtests as its



college placement assessment instruments. The correlation between GED Total score and



ACT Composite score was high with Pearson’s r = .80. There was no significant



difference between the ACT and GED relative to predicting performance in the English



composition class. Other studies were cited in this work that indicated the GED was



highly correlated to other standardized tests commonly used by colleges such as the SAT



subtests (Smith & Goetz, 1988). Although the GED subtests have changed since this



study was conducted, it still serves to illustrate the value of the concept of using the GED



as a placement test in order to reduce college expenses for GED completers (Smith &



Goetz, 1988).



Contrastingly, Rose (1999) suggested that the GED is not a good predictor of



college success. In a study conducted at a small four-year institution, 251 GED



completers had their ACT and GED scores correlated to their GPA for the 1997 spring



semester. Results indicated the ACT was a reliable predictor of college success while the



version of the GED Tests in use at the time was not a significant predictor of college



success (Rose, 1999).

27

Overview of Test Score Relationships



Standardized test scores are commonly used by institutions of higher education to



determine if students are prepared to be successful in their chosen course of study and to



place students in appropriate coursework. Institutions use a variety of tests and their



scores to determine the level at which students may be admitted to college or to certify



them for a particular profession or licensure. Nationally, test scores are used to determine



public policy and the efficacy of educational institutions and whether or not they need



improvement. Such tests necessarily must be given on multiple dates, in multiple



locations, and under multiple circumstances. To reduce the amount of testing error due to



students becoming familiar with the test and its items, multiple versions of each test must



be created. Multiple versions of a test however, create an additional testing consideration



namely whether the different versions of the test are equal in difficulty, constructs



measured, and content. Test score equating, scaling, and linking are all statistical methods



used by test makers and test users to make the scores of the various tests and their



different versions more useful (Kolen & Brennan, 2004).



Test Reliability and Validity



Reliability and validity are both critical features of any test that is worthy of use.



Test reliability refers to the extent to which the scores on a test reflect accurately what is



being measured. For a test to be reliable its scores must be consistent over time, reflecting



that a minimum of testing error is occurring. Internal consistency for test reliability can



be measured three ways. First, split-half reliability is a technique where the items of a test



are divided into two tests each containing one half of the items of the original test. If



when the two tests are administered and scored separately they result in two scores that

28

are consistent with each other, then the test has demonstrated internal reliability. The



second type of reliability is test-retest reliability or reliability over time. Test-retest



reliability is associated with an examinee’s consistency in responding in the same way to



the same item over time. If an item is vague or worded in a confusing manner it may



result in different responses by the examinee over time and thereby demonstrate that it is



possibly lacking in reliability (Sirkin, 2006).



Test validity can be defined as the extent to which a test actually measures the



concepts and content that it is meant to measure. There are many facets that impact on the



validity of a test. Face validity is the extent to which knowledgeable individuals agree



that a test meets its expectation for measuring a group of concepts or body of content. For



example, do the items on a fifth grade mathematics test appear to be applicable to the



kinds of mathematics taught to fifth graders? Content validity is the extent to which a test



covers the accepted range of a body of knowledge. Extending the example just used, does



that same fifth grade mathematics test adequately cover all the mathematics objectives



that fifth graders are expected to learn? Criterion validity is the extent to which a test has



the capability to predict accurately some criterion that is external to it. For instance, how



well do occupational aptitude tests predict the eventual occupational choice of



examinees? Finally, construct validity is the extent to which a test measures variables that



are related to a scale that the test is intended to measure. An example that describes



construct validity would be a test that measures personal satisfaction. A person who has a



high degree of personal satisfaction would be likely not to have an addictive or abusive



personality. It would be possible to empirically test and determine the degree to which a



person with a high score on the personal satisfaction scale exhibited addictive or abusive

29

behaviors. If these behaviors were absent or rare when compared to a high personal



satisfaction score, then the test would have established itself as having construct validity



(Sirkin, 2006).



Test Score Equating



Test equating is a statistical process that is used to adjust test scores from different



forms of the same test so that the scores can be used interchangeably. Test equating is



most useful when the difficulty and content of the different test forms are very similar.



Equating can adjust for differences in test difficulty but it does not adjust for differences



in test content. The purpose of test equating is to ensure regardless of which form of a



particular test is administered to a student, the student’s resulting test score has the same



meaning relative to the constructs being tested. In other words, it should make no



difference to a student which version of a test they attempt if the different forms are



identical or equal in difficulty (Kolen & Brennan, 2004).



Test Score Scaling



Scaling is a process that is related to test score equating and is used for comparing



scores from dissimilar tests. For instance, vertical scaling is used to compare



developmental scores or grade level equivalents for students from different grade levels.



Because these kinds of tests measure content that is matched to a specific grade level,



their scores can not be used interchangeably. However, a successful scaling of two



dissimilar test forms would produce a situation whereby scale scores could be used to



correspond between the raw scores of two test forms. For example, a test maker may



choose to establish a scale of 1-12 to correspond to the raw scores that can be received on



several different tests measuring the content associated with different grade levels. While

30

each grade level specific test may have twenty-five items, different raw scores on each of



the tests could indicate the same grade level achievement by an examinee. For instance, a



raw score of 20 on the test for first graders may indicate a third grade level of



achievement resulting in a scale score of 3. At the same time, a raw score of 10 on the



test for fifth graders may result in a scale score of 3, also indicating a third grade level of



achievement. In this way, scaling allows for the comparison of achievement results



between tests that are designed to measure different grade level specific content. The



differences in difficulty and grade level specific content of the various tests are



responsible for the different raw scores that correspond to the same scale score. If the



scaling process is successfully conducted, the scale scores from various test forms with



different levels of difficulty but which are measuring similar content can be used



interchangeably. In other words, the scale score of 3 on the first grade level and fifth



grade level tests in the preceding example would have the same meaning and value in the



way they describe students’ ability levels (Kolen & Brennan, 2004).



Test Score Linking



While test equating is used to establish the degree to which different forms of the



same test are identical in difficulty, construct measurement, reliability, and validity, test



score linking refers to the comparability of the scores from completely different tests



statistically. Score linking attempts to put the scores from two or more tests on the same



scale in a way that makes sense. Since it is unlikely that linking two different tests could



reach the level of sameness as equating two forms of the same test, linking should be



discussed in terms of adequacy. Different tests by virtue of being different must vary in



the way they are constructed and the content that they measure. While test linking can be

31

used to compare the relationships of the scores from different tests, the decisions and



processes used to link the scores bear mightily on the adequacy of those relationships.



Linking the scores from different tests can never be completely adequate in all instances



and for all groups and subgroups. Linking leans heavily on the expertise of informed test



makers and test administrators to make informed judgments about the adequacy of the



linkages between the scores of various tests (Kolen & Brennan, 2004).



Kolen and Brennan (2004) discuss three perspectives on test linking. First, linking



can focus on domains assessed by a test and the process of how a test was developed. The



domain of a test refers to (1) the framework definition or delineation of what the test will



measure, (2) the test blueprint or the mix of items, their formats, number of items, the



scoring rules, and other considerations, and (3) item selection or how well selected items



represent the test specifications. From this first perspective, how score linking can be



accomplished is based on how different and how similar the tests are relative to their



basic framework and specifications. Tests with similar framework and specifications are



naturally more capable of having their scores meaningfully linked. Tests with the same



framework but different specifications are likely to result in strong linkages because they



are measuring the same thing but in different ways. Tests with different frameworks and



different specifications are the least likely to have meaningful linkages established among



their scores for generalized populations. They can however, successfully be used to



establish meaningful score linkages for specific populations and for specific institutions



although those linkages are only meaningful over limited time intervals (Kolen &



Brennan, 2004).

32

A second perspective on score linking comes from two researchers Mislevy and



Linn, who have proposed four forms of test score linking based on the strengths of the



resulting linkages.



1. Equating. This is the strongest form of linkage and is defined as being



invariate across subpopulations. It is used to compare the scores from different



forms of the same test to determine if they may be used interchangeably and



yield identical results (Linn, 1993; Mislevy, 1992).



2. Calibration. This form of linkage uses statistical methods similar to equating



but is not population invariate. Calibration may refer to the relationship



between tests using the same framework but different test specifications. For



instance, test length affects the reliability of a test. All other things being



equal, a longer form of a test will be more reliable than a shorter form. As a



result, less able students would likely perform worse on a longer form of the



test than a shorter form and thus the forms are not population invariate for



linking purposes. Calibration is also appropriate for tests that have multiple



forms designed for measuring grade level specific achievement. These kinds



of tests have different content specifications and perhaps even different



statistical specifications. Finally, tests that employ an item response model



where all items in the domain are on the same common scale more



appropriately use calibration to determine relationships among different tests.



In the item response model, theoretically any subset of items that meets the



model’s assumptions of proficiency can be compared to a subset of items of



the same number (Linn, 1993; Mislevy, 1992).

33

3. Projection. This is a unidirectional form of linkage where scores from one



test predict the scores from another. Score linkages derived from projections



however are not reciprocal. That is, the best score projection to a second test



using a scale score from a first test may not result in the same score when



projecting backwards from the second test to the first. In this case, the tests



being compared can be different in constructs, content, and the domains being



measured. This form of linkage requires the use of a single group design



model to conduct a linking study. The projected relationship is almost always



obtained using a linear or non-linear regression statistic (Linn, 1993; Mislevy,



1992).



4. Moderation. This type of linkage takes two forms, statistical moderation and



judgmental moderation. Sometimes called distribution matching, moderation



usually employs the single group design where the same examinees take both



tests that are being considered for linkage. However, random group designs



and nonequivalent group designs are also possible. Concordance relationships



that result from moderation studies typically link tests with different



frameworks but similar constructs. Another kind of linkage study that uses



single group design determines the group mean and standard deviation for the



group’s score on two different tests and then adjusts the two tests’ scales to



have a common mean and standard deviation. The scores resulting from this



method will not be equatable. However, equal scores will not indicate an



equivalent level of proficiency as measured by the respective tests but indicate



only the likelihood of an examinee obtaining those scores if the examinee took

34

both tests. This same method of comparing standard deviations and group



means can be used with nonequivalent group studies as well and result in



levels of scores that are comparable. A more complicated form of moderation



can determine the relationship between the scores of tests with completely



different content and specifications such as American History compared to



Biology to compare the level of achievement and academic ability. This kind



of comparison requires the use of a moderator test to determine if the students



taking the American History test and the students taking the Biology test are



equal in general academic ability. Judgment moderation is the result of



informed experts making decisions about the scores that are required for



students to be determined as proficient on various tests. While based on



empirical data relative to examinee achievement, proficiency may be based on



a variety of other variables including scholarship dollars or available space in



a program, class or school. In this kind of score linking, proficiency may



change periodically and an acceptable score in one instance may be



unacceptable at another time. Likewise an acceptable level of achievement in



one area may be considerably different from acceptable score in a more



competitive area (Linn, 1993; Mislevy, 1992).



Linking studies may use a variety of designs to determine the association between



the scores of two tests. Generally, if a linking determination study uses a single group



design to collect data; a correlation coefficient is used to describe the strength of the



linkage. However, other types of designs such as random group designs and



nonequivalent group designs can be employed as well (Kolen & Brennan, 2004).

35

Kolen and Brennan (2004) discuss the third and final perspective on score linking



and look at score linking in terms of test similarity features or commonalities. While the



scores of any test can be linked, the utility and reasonableness of that linkage is largely



dependent upon the degree to which the tests share common features. Kolen and Brennan



suggest that there are four features to examine when determining the similarity of two



tests: inferences, constructs, populations, and measurement characteristics and conditions.



1. Inferences refer to the extent to which the scores of the two tests are used to



draw similar types of inferences and the extent to which the tests share



common measurement goals.



2. Constructs refers to the extent to which the two tests measure common



constructs such as higher level thinking skills including synthesis and analysis



as opposed to less complex thought processes like identification, recall or



recognition.



3. Populations refer to the extent to which two tests are designed to be used with



the same populations.



4. Measurement characteristics and conditions refer to the extent to which the two



tests share common measurement characteristics such as test length, test



format, and administrative conditions.



The inclusion of inferences sets this model apart from the model described by Mislevy



and Linn. The impact of inference on a linkage study is felt particularly in instances



where the test scores being linked are from tests that have been developed for distinctly



different purposes. The purpose for which the tests were developed may include the



relative stakes associated with the tests’ results. Linking scores from a high stakes test to

36

the scores from a low stakes test will likely result in different outcomes regarding the



relationship of their scores than linking scores from two low stakes tests or two high



stakes tests (Kolen & Brennan, 2004).



Dorans and Holland (2000) put forth the following descriptions of criteria that



they describe as critical to linking or equating test scores regardless of the perspective



being practiced.



Linking Criteria One – Similar Test Constructs: Test constructs refer to the



content and wording used in various test items and questions. Determining whether or not



tests have similar constructs and are measuring the same thing is accomplished through a



process of judging and comparing the content and wording of the two tests’ questions.



Tests that meet these criteria most often use the same test construction specifications and



blueprint. Tests that do not possess these commonalities of construction specification and



blueprint can only be linked by comparing test content based on carefully crafted content



definitions (Dorans & Holland, 2000).



Linking Criteria Two – Similar Test Reliability: Test reliability refers the degree



to which a test's results are consistent. To determine if two tests meet these criteria for



linkage, their respective reliability scores are calculated using standard methods and are



then compared. The closer the reliability scores of the tests, the greater degree to which



the tests might be linked in a meaningful way if they are also measuring similar



constructs and content (Dorans & Holland, 2000).



All of the perspectives on test linking described reveal clearly there is no one best



method or perspective for comparing the scores of two different tests. The purpose of the



different perspectives on linking studies seeks not to identify the sole best method but to

37

engage the researcher in exploring alternative ways to frame how their results should be



determined, used and interpreted (Kolen & Brennan, 2004).



The GED Tests and the COMPASS Tests



The General Education Development (GED) and the Computer-Adaptive



Placement and Support System (COMPASS) Tests have been in wide use for many years



in educational settings. As a result, it is very common for college applicants who are



GED completers to have taken both tests and possess scores that can be used for



comparative purposes. A content analysis that identifies the similarities of the two tests is



a necessary first exercise to determine the usefulness of a study that links the scores of



the two tests (Kolen & Brennan, 2004). The following description of the two tests under



consideration describes those tests’ history, reliability, validity, content, and constructs.



Overview of the GED Tests



The current English version of the GED Tests used in the United States has only



two forms available. As a result, stringent test security measures are in effect to maintain



the integrity of those test forms. Those forms are available for administration in this



country in a paper and pencil format only. The GED Tests were developed as a means of



determining the educational level of an examinee and were not intended for use as



diagnostic instruments or to determine college readiness. Regardless however, both



employers and institutions of higher education accept the GED as a credential that is



equivalent to a high school diploma. The GED Test Battery is composed of five tests:



Language Arts, Reading; Language Arts, Writing; Mathematics; Science; and Social



Studies. Each test has a maximum score of 800. Except for the writing sample, each test



is made up of multiple choice items, each with five possible answers (ACT, 2006).

38

The GED Tests: History



The GED Tests were developed in 1942 to measure the general concepts and



proficiencies associated with a high school diploma of that time. The tests were initiated



by the U.S. military establishment to assist veterans returning from World War II with



determining their future vocational and education goals. Originally, the tests were



administered to military personnel only. As it became apparent that there were wider uses



for the GED, the American Council on Education (ACE) undertook to capitalize on that



demand and began administering the test to civilians in 1952. The Veteran’s Testing



Service administered the test from 1942-1963 until its name was changed to the GED



Testing Service (GEDTS) in 1963. That name change was precipitated by the fact that



more civilians were being administered the tests than were former military personnel.



(GEDTS, 2007).



Annually, more than 875,000 examinees take the GED Tests worldwide at more



than 3,000 certified GED Testing Centers. To be certified, each center must meet



stringent requirements for test security and adhere strictly to the test material handling,



management, and storage protocols established by GEDTS. Testing services continue to



be available for military personnel stationed domestically and overseas along with U.S.



civilians and foreign nationals both here and abroad. Testing services are also available at



corrections facilities and certain health institutions. The test is accepted as the basis for



awarding high school equivalency credentials in all 50 U.S states and its territories as



well as all 11 Canadian provinces and territories. The test is available in U.S and



Canadian English versions, in Spanish, and in French. The English versions are available



in Braille, large print, and audiocassette (GEDTS, 2007). The purpose of the GED

39

according to the GED Testing service is, “...to provide an opportunity for adults who



have not graduated from high school to earn a high school level educational diploma. The



GED Tests measure the major academic skills and knowledge associated with a high



school program of study, with increased emphasis on workplace and higher education ”



(GEDTS, 2007).



The current version of the GED Test, the GED 2002 Series, began development in



1997. After field testing in 2001, each version of the test was standardized and equated



using a national sample of high school graduates. The test was released in its final form



in January of 2002. Test items underwent evaluation for difficulty, item discrimination



indices, and differential item functioning. Any items where less than 40% of examinees



answered correctly were eliminated from the item pool. Items with a point biserial



relation of less than 0.20 were also eliminated from the pool of eligible items. Each



resulting test form was constructed to have an average item difficulty of 0.70 and an



average discrimination index (average point biserial correlation) of 0.40. Items also



underwent two fairness procedures, judgmental sensitivity review, and differential item



functioning (DIF) screening. In the item sensitivity review, GEDTS staff members



identified and removed any items that might have been potentially offensive,



advantageous or disadvantageous to various groups of potential examinees. A DIF



analysis was also conducted on the tests’ items to determine if different groups perform



similarly on test items. Only items that met the standards for content and statistical



validity, have passed the sensitivity and DIF reviews, and possess appropriate levels of



item discrimination and difficulty were included in the final versions of the GED Test



(GEDTS, 2007).

40

The current series of the GED Test was normed on a representative sample of



graduating high school seniors who were given the GED Tests during March, April, and



May of 2001. Norming studies have been conducted for the GED Tests whenever it was



suspected that the achievement level of the norm group, graduating high school seniors,



may have changed or if the content of the GED Test itself changes (GEDTS, 2007).



To verify its test construction assumptions, the GED Test Service employs



Classical Test Theory (CTT) which is based on the concept that an examinee’s score on a



test is viewed as random sample of any number of possible scores that the examinee



could have earned on the test by taking the test or its parallel forms repeatedly. The



examinee’s score is viewed as being made up of the sum of a true score and a random



error component. Item Response Theory (IRT) on the other hand, considers the



examinee’s response to specific items as a predictable relationship based on the



examinee’s ability level. The GED Test Service considered switching to the IRT model



for its development of the GED 2002 Test Series but rejected it because of cost benefit



considerations even though it would have provided more information about examinees’



ability levels (GEDTS, 2007).



Overview of the COMPASS Tests



The ACT Computer Adaptive Assessment and Support System (COMPASS) is a



comprehensive assessment and diagnostic instrument developed to help postsecondary



institutions retain students by accurately advising and placing them in appropriate



coursework. The COMPASS is designed to assess the reading, mathematics, and writing



skills of students entering college. It may also be used to track academic growth and



diagnose specific academic deficiencies and proficiencies. All of the content areas of the

41

COMPASS are administered on computer including the writing sample. Each time the



test is administered, a new set of items are selected from an item pool automatically. This



feature practically eliminates the occurrence of duplicate items for individual examinees



who may take the test multiple times (ACT, 2006).



The COMPASS Tests: History



The COMPASS Reading, Writing, and Mathematics Placement Tests were



developed between 1985 and 1989 by establishing basic test specifications. Beginning in



1990, the ACT advisory panel for the COMPASS tests determined the technical and



content aspects of the tests’ development. The panel was divided into three work groups,



one for each of the three content areas. Each work group was responsible for developing



test items for its respective content area. In their subsequent meetings with college



faculty, counselors, and testing staff, the work groups determined that the system of



COMPASS Tests would be designed to act as a placement and diagnostic tool for its



users as well as provide them with a variety of supplemental statistical information and



reports (ACT, 2006).



Test items for each of the content areas were developed in the same manner. First,



the work groups conducted literature reviews to determine what kind of content was



relevant to the applicable entry-level courses at two and four-year postsecondary



institutions. The work groups also reviewed samples from the course catalogues of 23



institutions across the nation to gather additional information about entry-level, remedial,



and advanced course content. As test items were developed and submitted for inclusion in



the test item pool, ACT conducted both an internal and external review of the items. The



internal review consisted of a review of the items by ACT staff for fairness, content

42

accuracy, and general quality. If an item was determined as needing revision, it



underwent the same review after being resubmitted in its then revised form as it did



originally. Test items were inspected to ensure that there was only one correct response,



that distracter items were plausible but incorrect, and that the distracter responses were of



an appropriate cognitive level. All test materials were reviewed for sexist language, fair



portrayal, and balanced representation of various societal groups (ACT, 2006).



An external review of the tests’ item pool was conducted by a series of



consultants engaged by ACT in 1992. These external reviewers represented five groups,



African-Americans, Asian Americans, Latino/Latina Americans, Native Americans, and



Women. A fairness panel having a representative from each group was constituted from



individuals selected by ACT. The panelists were selected from lists of names provided to



ACT by nationally recognized advocacy groups. Each panelist was provided with the



items suggested for inclusion in the COMPASS Test item pool. After reviewing the



items, the panelists participated as a group in a teleconference with ACT staff and each



item was discussed. Following the teleconference, ACT staff met to discuss each



panelist’s comments and determined whether to eliminate, revise or leave items



unchanged.



In 1997, five panels were convened for the same purpose as the original panel.



These five fairness panels again represented each of the five groups with each panel



representing a specific focus: African Americans, Asian Americans, Latino/Latina



Americans, Native Americans, and Women. In total, twenty-five panelists were engaged



for this second external review. Participants for this review were academicians who were



identified as being sensitive to issues affecting the five groups’ perspectives through their

43

experience in teaching, advocacy, or mentorship of students in those respective target



populations. This process of internal and external review of test items remains in effect



for all new items developed for inclusion in the COMPASS Test item pool (ACT, 2006).



The review of the literature points to the growing importance of accountability to



postsecondary institutions as measured by the college success of their students. It also



connects and reveals the variety of factors that influence college readiness and college



success. Finally, the literature review discusses test equating and test score linkage



practices along with the characteristics of the COMPASS and GED Tests. To explain the



processes of how the research questions stated earlier will be answered, the methodology



section that follows will describe the techniques for conducting a content analysis,



correlation, linear regression analysis and equipercentile scaling in the comparison of



tests and test scores.

CHAPTER THREE



METHODOLOGY



The purpose of this study was to determine whether GED Reading and



Mathematics Test scores could be used to accurately predict scores on the COMPASS



Reading and Mathematics Placement Tests respectively. To accomplish its purpose, the



study sought the answers to these five research questions regarding the GED and



COMPASS Tests:



1. To what degree are the tests measuring the same content?



2. To what degree are the tests similarly reliable?



3. To what degree are the tests symmetrical?



4. To what degree are the tests’ scores linkable?



5. To what degree are any test score linkages population invariant?



Research Design



The research design for this study employed three analytical methods. First, a



content analysis was conducted comparing the two tests to determine the degree to which



they were measuring the same contents and were similarly reliable. The results from the



content analysis were used to answer the first and second research questions. Second,



since the tests were found to be sufficiently similar in the contents they measure and in



reliability, a correlation and linear regression analysis was conducted using scores from



study subjects who had taken both the GED and COMPASS Tests. This analysis was



conducted to determine the strength of the relationship between the scores of the two tests



overall. The results of the correlation and linear regression analysis provided answers to



the third, fourth and fifth research questions posed for this study. Finally, for ease of

45

comparison the tests’ scores were placed in concordance tables respectively using



equipercentile scaling methods.



Content Analysis



Instrumentation



To facilitate the content analysis of the various tests under consideration, two



methods were employed. First, a narrative description of the contents measured by both



tests was completed to give a sense of the commonality of what the tests were trying to



assess. Second, a set of tables that set out the technical aspects of the COMPASS



Reading and Mathematics Placement Tests and the technical aspects of the GED Reading



and Mathematics Tests were completed to provide ease of comparison. The technical



aspects of the tests being analyzed for this study included: reliability score, validity score,



total number of items per test, time allowed for completing test, score ranges, item



configuration, and testing formats available, and whether or not calculator use was



acceptable during the two tests’ mathematics portion. These technical aspects were



identified and listed on a set of analysis tables and the information relative to each item



was recorded for both the GED and COMPASS Tests. A second set of tables was created



to facilitate the comparison of the contents measured by the COMPASS and GED Tests



to identify and compare their respective content descriptors.



Content Analysis Procedure



The documentation reviewed for this content analysis of the GED Reading and



Mathematics Tests and COMPASS Reading and Mathematics Placement Tests consisted



of the tests’ respective technical manuals, selected test preparation study guides, and



practice tests. For security reasons, actual copies of the GED examinations were

46

unavailable for review. Likewise, the actual COMPASS Tests were not available for



content analysis because of test security considerations. However, both the COMPASS



and GED Tests have published technical manuals that described in detail the tests’



purpose, constructs, content, reliability, and validity. These sources provided ample data



for the comparison of these two instruments’ similarities and differences.



To create the content analysis comparison tables, the content areas described for



the COMPASS Test in the applicable technical materials were reviewed and entered on



the tables. The list of content areas were derived from the COMPASS Test Technical



Manual and included content items for each of the tests under consideration (ACT, 2006).



Following that process, the GED content descriptions that closely matched those



described in the COMPASS technical documents were identified from the GED technical



materials available for review and entered into the tables. In each instance, the language



used by the two tests was carefully interpreted to determine when content was being



described in similar language and when the same content was being described using



alternate terminology.



Overall, the material that was associated with each test was thoroughly analyzed



to determine the degree of similarity that each test’s constructs and content exhibited to



its counterpart. The resulting descriptions and tables served to provide a clear comparison



of the differences and similarities of the two tests’ content, constructs, and reliability and



gave ample evidence regarding their usefulness in a score linkage study. According to



Kolen and Brennan (2004), the greater the similarity of the characteristics between tests,



the greater the likelihood that their scores could be meaningfully linked (Kolen &



Brennan, 2004).

47

The completion of a content analysis was the first step in determining if scores



from the two tests could be linked in a meaningful way. The documents described



provided an adequate sampling of the content, constructs, and reliability characteristics of



the applicable GED and COMPASS Tests necessary to determine if the tests were



sufficiently similar. The content analysis results confirmed affirmatively that the tests



were sufficiently similar to warrant further exploration of the strength of the relationships



between the tests’ scores through a correlation and linear regression analysis. The



resulting linear regression analysis indicated that the two test’s scores were related to a



significant degree and opened the way for the creation of concordance tables using the



equipercentile scaling method. The resulting concordance tables then were constructed to



gain insight into the value of using GED Test scores to predict COMPASS Test scores.



Test Score Data Analysis



Study Subjects



The test scores included in this study were from students who were enrolled in



semester credit courses at Houston Community College during the 2006 calendar year



and who had completed the GED Tests between the dates of January 1, 2002 and



December 31, 2006. These parameters were chosen for selection of the study’s data sets



for these reasons: (1) On January 1, 2002, the version of the GED Tests currently in use



was put into service (GEDTS, 2007). Therefore, GED completers who had tested using



previous versions of the test could not be included because the tests they had completed



were not comparable to the current version of the test. (2) The scores of students enrolled



in semester credit courses during the 2006 calendar year were selected for the data set



because they represented the most recently available full year’s worth of scores from

48

GED completers who had taken the current version of the test. Completers from the 2007



calendar year were not included because of a prequalification pilot project being



conducted at HCC by the State GED Chief Examiner’s Office during that year.



Candidates for the GED during the period of the pilot project were required to pass a



prequalification test before attempting the actual GED battery of exams. Those GED



completers were excluded from this study because their test scores were likely to be



higher on average than those from non-prequalified completers. Only the scores from



GED completers who were enrolled in semester credit hour courses were included in the



data set because other course offerings at the college in which GED completers might



have enrolled may not require a COMPASS test for admission purposes.



In addition to the data sets to be collected, the characteristics of the study subjects



were descriptively represented in a series of charts included in the data analysis section of



the study. Those characteristics included: ethnicity, gender, age at GED completion, and



age at withdrawal from school. Study subject characteristics selected for description to



provide important information about the group that may have influenced the results of the



study, especially the age of the subjects when they completed the GED and their age



when they left school. These age-related characteristics were selected for description



because of their influence on academic performance of students. The literature



consistently confirmed that staying in school longer generally increased the amount of



academic preparation students obtained and then consequently made them older at



school exit (CALEC, 1995; Miglietti & Strange, 1998).

49

Test Score Data



Data collected and analyzed for the study included COMPASS Test scores for



Reading and Algebra along with GED Reading Test scores, GED Mathematics Test



scores, and GED Test score averages. Also included in the data gathered for the study



were the subject’s age, ethnicity, grade at time of school exit, percentile rank for GED



Reading Test score, and percentile rank for GED Mathematics Test score. Percentile rank



for GED Test scores in this instance referred to the students’ predicted high school rank



in class national averages based their GED Test scores (GEDTS, 2007).



Following approval by the University of Houston Committee for the Protection of



Human Subjects, a request was made to Houston Community College for the acquisition



of COMPASS Test score records from the 2006 calendar year for all GED completers



who had been enrolled at the institution in semester credit hour course work during that



year. At the same time, a request was made to the Houston Community College GED



Chief Examiner’s Office for the acquisition of the GED Test score records for all students



contained on the list of 2006 calendar year GED completers enrolled in semester credit



hour coursework. After approval by the Houston Community College administration



responsible for student records and obtaining both sets of information, the lists were



compared and only those GED completers with GED Test scores occurring after January



1, 2002 and before December 31, 2006 were retained for the study group. The selected



records were given sequential numerical designations and entered into SPSS for analysis.



All identifying fields for the study subjects were expunged following the development of



the final study subject list. All records with identifying information were scheduled to be



destroyed in accordance with the conditions stated on the application for the conduct of

50

research on human subjects approved by the University of Houston Committee on the



Protection of Human Subjects.



Data Analysis Procedure



Correlation



To determine the strength of the relationship between the GED Test scores and



COMPASS Test scores, a correlation and regression analysis was conducted on the data



sets. The results from this analysis were placed in chart form for ease of interpretation



and were included in the Presentation of Findings section along with a narrative



description of the other findings of the study. Before conducting the regression analysis, a



Coefficient of Correlation, or Pearson’s r was first generated to determine the strength of



the relationship between the two sets of data. If a strong relationship exists between the



data sets, it will be indicated by a large r suggesting that the two sets of data are likely to



have a meaningful relationship. In addition the r value for the set of data, the Coefficient



of Determination or r2 was also calculated. The Coefficient of Determination (r2)



indicated how much of the change in score on the COMPASS Tests was explained by the



changes in the GED Tests’ scores. Without a strong Coefficient of Correlation (r) and



Coefficient of Determination (r2), it would have been unlikely that the variables under



consideration were meaningfully related and there would have been little point in



conducting further analysis of their relationship. However, a strong or moderate strength



r and r2 value would suggest that the relationship between the GED and COMPASS Test



scores under consideration could be meaningful and that GED Test scores could be



predictive of COMPASS Test scores (Sirkin, 2006).

51

Regression Analysis



The purpose of the regression analysis was to produce a linear regression equation



that described the effect that changes in the value of the independent variable, GED Test



scores, had on the value of the dependent variable, COMPASS Test scores (Sirkin,



2006). The determination of a linear regression equation allowed the value of a dependent



variable to be predicted based on the value of an independent variable. The linear



regression model used for the purposes of this study was the Least Squares Method. This



method was useful for determining the regression equation from the study’s data points



because they did not represent a perfect line and produced an equation that most



accurately predicted the effect of changing the value of the independent variable had on



the value of a dependent variable. The linear regression equation derived using the Least



Squares Method described a line that would result when the deviations by individual data



points from a hypothetical line were squared individually and added together to result in



the smallest sum possible (Sirkin, 2006).



The linear regression equation for this study was determined by representing the



values of the independent variable, GED Test scores, along the x-axis of a Cartesian



plane, while the values of dependent variables, COMPASS Test scores, were represented



along the plane’s y-axis.



The line that best represented the relationship between the sets of data points in a



linear model could then be described using the formula ŷ = axy+byxx. In this formula, ŷ



results in a line that estimated the value of y based on the regression and correlation



calculations described earlier. The term axy represented the point at which the line formed



by the equation would intercept the y-axis when plotted on a Cartesian plane. The slope

52

of the regression line was represented by byx. The full formula for the linear regression



equation as described is:





y

 y  b  x  n xy   x  y 

yx



n x   x 

2

n 2







where y equals the value of the dependent variables, x equals the value of the



independent variables and n equals the number of cases examined. A regression line



equation based on this formula could then be used to estimate the value of y otherwise



written as ŷ when given the value of x (Sirkin, 2006).



Equipercentile Scaling



Equipercentile scaling was used as method of relating the scores from the



different tests for predictive purposes. This method was used to set the scores from the



different tests equal or near equal according to their percentile rank within their



respective data sets (Laverge & Walker, 2006; Schneider & Dorans, 1999). For example,



if a score of 500 on the GED Reading Test was ranked at the 50th percentile among the



GED Reading Test scores of the study subjects, it was concorded to the score that also



ranked at the 50th percentile for COMPASS Reading Placement Test scores among the



study subjects.



To conduct the equipercentile scaling procedures, the scores for GED Reading,



GED Mathematics, COMPASS Reading and COMPASS Algebra Placement Test were



placed in percentile rank by test respectively using the scores for all study subjects. The



small sample number prevented the same procedure from being conducted for subgroups



by gender, ethnicity, and a combination of gender with ethnicity.

53

For the purposes of this study, COMPASS Test score percentile rankings were



aligned as closely as possible to GED Test score percentile rankings within a minimum of



2 percentile points and in no case were they larger than the concordant GED Test score



percentile ranking. To be conservative in construction of the concordance relationships,



in cases where scores did not meet the percentile ranking alignment criteria established



for this study, applicable COMPASS Test scores were always be placed in concordance



with the next lowest GED Test score available.



In summary, the research design of this study included both qualitative and



quantitative methods. Content analysis, a qualitative method, was used to determine the



relative similarities and differences of the content, constructs measured, reliability, and



validity of the GED and COMPASS tests under consideration. Correlation and linear



regression analysis along with equipercentile scaling, both quantitative methods, were



employed to determine the strength of the relationships between GED Tests scores and



COMPASS Test scores and created a framework for prediction of COMPASS Test scores



from GED Test scores. In combination, the results of the methods used in this research



study provided valuable insights into prediction of college readiness for GED completers.



Study Limitations



This study is limited in its scope because it looks only at the degree to which the



Reading and Math portions of two tests, COMPASS and GED, were linkable. Even when



instruments being compared are similar, many other factors can influence the college



readiness of students and the predictive value of one test score for another. It is also



limited because it looks only a single measure for predicting college readiness when



clearly the literature suggests that college readiness and college success are influenced by

54

a variety of factors. For instance, the point at which a student left school could have an



influence on their overall academic abilities and this could affect the student’s GED



score. As the literature suggests, the longer students stay in school, the more likely they



are to experience opportunities to improve academically (CALEC, 1995). This study



likewise does not consider the high school Grade Point Average (GPA) of the GED



completers before they left school. Strong academic performance and prior achievement



in school are clearly predictors of strong academic performance in college and have a



stronger correlation in many cases than do standardized test scores (CALEC, 1995;



Garavalia & Gredler, 2002; Naumann, Bandalos, & Gutkin, 2003; Popham, 2006). This



kind of relationship could also be true for GED completers. Their age along with how



long GED completers have been away from formal schooling is also not a consideration



in this study. These two variables, however, are clearly influential in the performance of



students in college. Students’ belief in their ability to successfully perform in college or



student self-efficacy is also positively correlated to college success and is likely to be



influential for GED completers in college as well (Golden, 2003; Lamkin, 2004).



This study however, does not take into consideration the self-efficacy of GED



completers regarding their belief in their own ability to be successful in college.



Additionally, other student characteristics that impact college readiness and success like



work ethic, social support network, and other personal and personality characteristics



(Ridgell & Lounsbury, 2004) are likewise not considered as factors in this study. Neither



does it take into consideration institutional factors that can affect student success and



retention (Lau, 2003). Finally, this study does not consider any kinds of preparation



classes which students might have accessed prior to enrolling in college or attempting

55

college placement assessments. These kind of preparations clearly influence college



placement scores (Stovall, 2000) and thereby may affect the consistency of GED scores



as a predictors of college readiness.



The study is also limited because the actual test instruments themselves are not



available for review due to security considerations. However, the technical manuals and



test preparation materials adequately describe and define the content, constructs, and



reliability of both the GED and COMPASS Tests, and ultimately result in a credible



analysis of where the tests possess similarities and where they differ.



Finally, study is limited because its sample size resulted in subgroup numbers that



were too small for correlation calculations or regression analysis. For instance, only four



students identified themselves as Asian, one male and three female. While only eleven



students identified as White, six male and five female and only five students identified as



African-American males were included in the study.



In addition to the limitations of sample size, the test data collected for this study



came only from students who had first taken the GED Tests and had then subsequently



had taken the COMPASS Tests. This situation prevented the analysis of the symmetry of



the two tests. For a symmetry analysis to have been conducted, reciprocal data from



students who had first taken the COMPASS Tests and then subsequently taken the GED



Tests would have had to have been available for analysis and comparison to the



relationships identified in this study. If the respective strengths of the relationships found



for the two groups were found to be similar, then the tests would likely exhibit some



measure of symmetry (Kolen & Brennan, 2004). The next section describes the findings



of the study.

CHAPTER FOUR



PRESENTATION OF FINDINGS



The findings of the study are organized into two sections. The first section is a



description in narrative and chart form of the findings derived from the content analysis



of materials associated with GED Tests and COMPASS Tests. The intent of the content



analysis is to explore the tests’ similarity in regard to their content assessed and



reliability scores. While differences in test reliability scores can readily be ascertained,



determining the degree to which different tests are measuring similar contents can be



difficult. The common methodology employed for that purpose consists of inspection of



the content of the tests and how the tests’ items are worded (Dorans & Holland, 2000).



The analysis of these two test characteristics is essential to answering research questions



one and two for the study:



1. To what degree are the tests measuring the similar content?



2. To what degree are the tests similarly reliable?



The second section starts out with a description of the characteristics of the



subjects whose GED and COMPASS Tests scores were used in the study. Following the



description of the study subjects, the results of a linear regression analysis of GED and



COMPASS Test scores is presented. The intent of the regression analysis is to explore



the degree to which the GED and COMPASS Tests are symmetrical, the degree to



which their scores are linkable, and the degree to which any scores linkages are



population invariant. The analysis of these characteristics is necessary to answer



research questions three, four, and five for the study:



3. To what degree are the tests symmetrical?

57

4. To what degree are the tests’ scores linkable?



5. To what degree are any test score linkages population invariant?



Finally, this section will conclude with a set of concordance tables that resulted



from an equipercentile scaling procedure comparing GED and COMPASS Test scores.



A summary of the findings and the analysis of how they answer the study’s research



questions and hypotheses will be contained in the Summary and Conclusion portion of



the study.



Content Analysis and Reliability Findings



The GED Tests



GED Documents



GED 2002 Transitions: A Guide for Chief Examiners



This document was obtained from the GED Chief Examiner’s Office of the



Houston Community College System. This manual described the protocols for opening



an official GED Test Center in Texas. Its content covered staffing requirements and



qualifications, official GED Test Center operating procedures, and GED Test



administration requirements and practices. Any entity in Texas certified to administer



the GED Test must have developed a manual of this nature for each GED Test Center



that it proposed to operate. The location of testing and the contents of the manual must



have been approved by the Texas Education Agency’s GED Chief Examiner’s Office



before any GED Tests were administered (HCCS, 2002).



GED Test Technical Manual



The GED Test Technical Manual was found on the GEDTS website.



Information on this website included access to the 2007 version of the technical manual

58

associated with the GED battery of tests. This document described the history, purpose,



test specifications, and development procedures followed for the creation of the current



version of the GED Tests. This most recent iteration of the GED Test Technical Manual



described thoroughly the tests’ content, context, format, and cognitive level (ACE,



2007a).



GED Complete Preparation



This document could be described as a preparation study guide and was



published by the Steck-Vaughn Company for individuals preparing to attempt the GED



Tests. It included sample questions presented in the same format that individuals would



experience them on the actual GED Tests. It also described how the tests were timed



and the skills that were assessed by each of the tests. The guidebook also discussed



other information helpful to test takers including how to prepare for the tests, test taking



skills, and study skills (Northcutt, 2002).



The Official GED Practice Test



The Steck-Vaughn Company has been exclusively licensed to distribute the



GED Official Practice Tests (OPT). The OPT, when administered correctly, it has been



designed to present individuals with items and conditions identical to those of the



official GED examinations. Scores on the OPT have been crafted to closely mirror those



that the individual might be expected to encounter on the official GED Tests (GEDTS,



2003a, 2003b, 2003c).



Domains of Assessment



The GED Tests cover five domains: Language Arts, Reading; Language Arts;



Writing, Social Studies; Science; and Mathematics. The highest score that may be

59

obtained on any individual test in the GED Battery of Examinations is 800. Examinees



that complete the entire battery receive both individual test scores as well as a total



composite score, which is the average of the scores of all five tests (Northcutt, 2002).



Examinees may attempt the tests multiple times but are required to wait at least 6



months before retesting unless approved for retesting by a recognized GED test



preparation program. The highest score obtained on each respective test is used in the



calculation of the examinee’s total composite score. To receive a GED credential in



Texas, a examinee must obtain an average score of at least 450 on all five GED Tests



with no single test score used in the calculation being less than 410 (HCCS, 2006). For



the purposes of this content analysis, only the GED Language Arts, Reading and GED



Mathematics Tests will be considered because the COMPASS Test has no equivalent



form for comparison to the GED Science or Social Sciences Tests. In addition, the GED



Writing Test and COMPASS Writing Test will not be considered as part of this analysis



because every study subject was not required to complete a writing sample for the



COMPASS Test (HCCS, 2004). In contrast, every GED examinee must complete a



writing sample as part of the GED Writing Test (Northcutt, 2002).



GED Reading Test Content



The GED Reading Test is a timed test lasting no longer than 65 minutes and



consists of 40 multiple choice questions. Each item has five choices from which the



examinees may select a single response. Examinees record their answers on optical



mark score sheets provided for them. The test analyzes the reading comprehension and



analytical ability of the examinees. Examinees must exercise five types of thinking



skills to answer the questions: comprehension, application, analysis, evaluation, and

60

synthesis. Reading selections for the test include two nonfiction selections, three prose



selections, one poetry selection, and one drama selection. Twenty-five percent of test



items relate to the nonfiction selections while the remaining 75% of test items relate to



the literary texts (Northcutt, 2002).



GED Mathematics Test Content



The GED Mathematics Test is a timed test lasting no longer than 90 minutes and



consists of 50 multiple choice questions. Like the GED Reading Test, each item has five



choices from which the examinees may select a single response and examinees likewise



record their answers on optical mark score sheets provided for them. The GED



Mathematics Test analyzes four content areas: Numbers and Operations; Measurement



and Data Analysis; Algebra; and Geometry and tests the constructs of comprehension,



application, analysis, evaluation, and synthesis (Northcutt, 2002).



About half of the test involves the use of drawings, diagrams, charts, and graphs.



Examinees use may a calculator issued by their respective GED Test Centers to help



answer those questions. Examinees are also issued a formulas page containing common



mathematical formulas that they may use during the test (Northcutt, 2002).



Between 20-30% of the items on the GED Mathematics Test relate to Numbers



and Operations. These items require the use of arithmetic operations to solve formal



mathematical problems and real world situations. A calculator may not be used for this



section of the test (Northcutt, 2002).



Items relating to Measurement and Data Analysis make up 20-30% of the GED



Mathematics Test. This section of the test requires the examinees to solve problems that



involve determining length, perimeter, volume, circumference, time, and other

61

measurement related questions. Data analysis is tested by how well the examinees use



charts, graphs, and tables. In this section, the examinees are also asked to calculate



averages, means, modes, and probabilities (Northcutt, 2002).



Algebra questions account for 20-30% of the GED Mathematics Test. The



examinees’ knowledge of algebraic concepts like variables, equations, square root,



exponents, and scientific notation are tested along with use of the coordinate plane for



the graphing of ordered pairs, equations, solving inequalities, and determining the slope



of a line. This section of the test may require the examinees to use an alternative answer



format by having them mark their answers on a coordinate grid (Northcutt, 2002).



Finally, Geometry concepts comprise 20-30% of the items on the GED



Mathematics Test. Examinees are asked to respond to items about lines, circles,



triangles, and quadrilaterals and use arithmetic operations to find values of angles and



line segments. The test also contains items that measure knowledge of the Pythagorean



Theorem and congruence relationships. The test measures no trigonometric content



(Northcutt, 2002).



Reliability



According to the literature reviewed, the reliability of the GED Tests’ multiple



choice items are evaluated three ways: (1) by determining the tests’ internal consistency



reliability, (2) by determining the standard error of measurement, and (3) by



determining the alternate form reliability of tests. All reliability analyses for all forms of



the GED Test utilize random samplings of high school graduates (GEDTS, 2007) .



The estimates of the internal consistency reliability of the GED Tests are based



on a K-R 20 reliability coefficient which ranges from 0-1. The K-R 20 statistic is an

62

estimate of the extent to which all of the items on a test correlate positively to one



another. It is also an estimate of the extent to which alternate test forms of the same



length correlate to the original version. The results of the equating studies conducted by



the GED Testing Service on all of the various forms of the GED 2002 test version



indicate that their K-R 20 reliability coefficients are at minimum .92, with over 80% of



the tests alternate forms having a K-R 20 of .94 or higher. This range is consistent with



other commercially available achievement tests (GEDTS, 2007).



The standard error of measurement (SEM) is an estimate of the amount of error



that is associated with the scores derived from a test. This statistic is used to describe



how far an examinee’s observed score differs on average from a score without error and



represents their true score. Because the SEM represents a confidence interval of one



standard deviation, an examinee’s observed scored is likely to fall within the predicted



range 68% of the time. SEM can not be compared across test forms without considering



the unit of measurement, range and standard deviation of raw test scores. The SEMs for



the entire set of GED 2002 test forms are typically within 25% of the various forms’



standard deviations. Tests with SEMs that are less than one third of their standard



deviations are considered acceptable when their reliability coefficients are 90% or



higher. The mean score for the various forms of the GED Reading Test ranges from



497.8 to 518.2 with a standard deviation score range of 121.4 to 134.3. The various



forms of the GED Mathematics Test have a mean score range of 477.9 to 526.7 with a



standard deviation score range of 101.0 to 126.9 (GEDTS, 2007).



Alternate form reliability refers to the correlation of the scores between different



forms of a test that are administered to the same groups of examinees. Since both forms

63

of the test are intended to measure the same proficiency, are developed from the same



content specifications, and are designed to have the same psychometric characteristics



they should exhibit strong similarity and the test forms should produce strong alternate



form reliability. A study conducted by the GEDTS in 2004 involving 77 schools and



2,557 graduating seniors obtained alternate reliability coefficient correlations ranging



between .70 and .83 (GEDTS, 2007).



Validity



Validity is the degree to which all of the accumulated evidence supports the



intended interpretation of test scores for their proposed purpose. Sources of evidence



that provide information relative to the validity of a test include test content, response



processes, internal structure, and relation to other variables. The purpose of the GED



Test is to measure the academic knowledge and skills achievement that would typically



be accumulated during a four year experience in high school (GEDTS, 2007).



To confirm the content validity of the tests, GEDTS used the standard practice of



relying on the subjective analysis and opinion of subject–matter experts. To ensure the



content-related validity of the GED Tests, nationally representative groups of experts



were used to develop test specifications and evaluate the final forms of the various tests



(GEDTS, 2007).



Evidence of test validity based on response processes was derived from a study



that correlated the scores of GED examinees who took the tests during 2002-2003. This



study resulted in findings that showed the various subtests to be related to each other



with correlation values ranging from .53 to .77 (GEDTS, 2007).

64

Evidence-based relations with other variables or criterion-related validity refer to



how well a test relates to other tests that measure the same or similar attributes. Studies



conducted by GEDTS indicate that pass rates for the sample of high school seniors



taking the GED Tests for standardization purposes and actual GED completers compare



favorably at 84% for the former and a range of 83-93% for the latter. Additionally, a



correlation study using the GED Test scores of high school seniors who participated in



the 2002 equating study and their self-reported high school grades was conducted and



found significant correlations for every test at the p < .001 level. The results of the study



indicated that higher grades in high school correlated to a higher likelihood of passing



the GED Test (GEDTS, 2007).



Clearly, the GED Testing Service practices an extensive process for maintaining,



monitoring and developing new test forms and the GED Tests meet a rigorous standard



for both reliability and validity. While the test is not designed to act a predictor of



academic success in postsecondary school or the workplace, its usefulness in measuring



academic skills and achievement in core content areas, cause employers and institutions



of higher education to readily accept it in place of a high school diploma (GEDTS,



2007).



The COMPASS Tests



COMPASS Documents



COMPASS/ESL Reference Manual



This document described the appropriate use of the COMPASS Tests. It



described how the tests were developed and how they could be used in placement

65

testing for college entry. The manual detailed the content, context, reliability, and



validity of the tests (ACT, 2006; GEDTS, 2007).



COMPASS Preparation Material



These documents discussed how to prepare students for the COMPASS Tests



and provided examples of the kinds of items that test takers could experience on the



actual test. These preparation guides were paper and pencil versions of the COMPASS



Mathematics and Reading Placement Tests which were otherwise computer-adapted.



These test examples do not produce a sample score but were intended to only familiarize



students with the formats of the real tests (ACT, 2004a, 2004b).



Domains of Assessment



The COMPASS Test assesses Reading, Mathematics and Writing. Because the



tests are computer adaptive, each test taken by a student will be made up of different



items. The maximum score on each of the tests is 99. The Mathematics Test is made up



of five separate tests: Numerical Skills/Prealgebra, Algebra, College Algebra,



Geometry, and Trigonometry. Scores from the COMPASS Writing Test were not



included in this study because, not all students were required to take the COMPASS



Writing Test for college readiness purposes (ACT, 2007; HCCS, 2004).



COMPASS Reading Placement Test Content



The COMPASS Reading Placement Test assesses an examinee’s reading skills



and determines if the examinee is prepared to successfully enter college level



coursework. The COMPASS Reading Placement Test concentrates on assessing the



examinee’s ability to construct meaning from what is read and therefore focuses on



reading comprehension. It does not measure vocabulary knowledge. There are five types

66

of passages in the COMPASS Reading Placement Test. (1) Prose fiction passages



emphasize the narration of events or revelation of character; (2) Humanities passages



describe or analyze ideas or works of art; (3) Social Science passages describe



information discovered through research; (4) Natural Science passages present scientific



information along with a discussion of its significance; and (5) Practical Reading



passages present text relative to vocational or technical courses. All passages are taken



from published materials or are original works written by item writers contracted for



that purpose by ACT. The passages are intended to reflect the content and rigor of



reading experiences that would confront a first-year college student. Each passage is



designed to be presented to the examinee with up to five multiple-choice items, each



having five options for a single correct response. These items are designed to determine



the examinee’s comprehension of the passages’ text, message, and meaning. The five



comprehension items that accompany each passage can be categorized as reasoning and



referring items. Referring items pose questions explicitly about information found in a



passage. Reasoning items test the examinee’s ability to make inferences, develop



understanding, and derive the meaning of unfamiliar or ambiguous words from the



passage (ACT, 2006).



COMPASS Mathematics Placement Test Content



The COMPASS Mathematics Placement Tests are developed around five



content domains: numerical skills and prealgebra, algebra, college algebra, geometry,



and trigonometry. Examinees can be tested in one more of these domains for placement



purposes and for diagnostic purposes in two of the domains. Each of the five content



domains has an item pool of about 200 multiple-choice items each with five response

67

options. Items can be categorized into three general levels of cognitive complexity. (1)



Basic skills items can be solved using a series of basic mathematic operations. (2)



Application items require the use of basic operations to new situations or in a complex



manner. (3) Analysis items require examinees to demonstrate understanding of



concepts, principles, and relationships that are relevant to the mathematical situations



introduced in each test item (ACT, 2006).



The COMPASS Numerical Skills and Prealgebra Placement Test is the most



basic of the five mathematic domain tests. Items in this test range from basic arithmetic



concepts using basic operations with integers, fractions, and decimals to the prerequisite



skills for entry into algebra such as exponents, absolute values, and percentages. This



test is generally used to determine whether or not an examinee should be placed into a



elementary college algebra class or into some lower level of remedial mathematics



(ACT, 2006). No score on the COMPASS Numerical Skills and Prealgebra Placement



Test, not even a perfect score, is considered an indication that an examinee is college



ready (HCCS, 2004).



The COMPASS Algebra Placement Test is made up of items from three areas of



the mathematics curriculum; elementary algebra, coordinate geometry, and intermediate



algebra. Examinees that score high on this test should be routed to further assessment



with the College Algebra Placement Test. Those examinees scoring poorly should be



routed to additional assessment with one of the COMPASS diagnostic tests for more



accurate placement (ACT, 2006). Students scoring 71 or higher on the COMPASS



Algebra Placement Test are considered college ready and capable of being successful in



freshman level algebra classes at Houston Community College (HCCS, 2004).

68

The COMPASS College Algebra Placement Test is most appropriate for



students who have performed well in intermediate algebra courses in high school. The



COMPASS College Algebra Test includes concepts like functions, exponents,



factorials, linear equations, and roots of polynomials. Examinees scoring low on the



COMPASS College Algebra Test should be routed to further assessment with the



Algebra Placement Test. Examinees scoring high on the COMPASS College Algebra



Test should be routed to the COMPASS Geometry or Trigonometry Placement Tests for



further assessment (ACT, 2006).



The COMPASS Geometry Placement Test assesses an examinee’s



comprehension of Euclidian geometry and their ability to use spatial reasoning and



geometric principles to solve problems. The content items in this test include those



relative to triangles, circles, angles, rectangles, and polygons as well as logic and proof



statements. Placement decisions should be based on scores from this test along with



other mathematics placement information (ACT, 2006).



The COMPASS Trigonometry Placement Test assesses an examinee’s



understanding of trigonometric functions and their application to solving problems. Test



items cover content including trigonometric functions and identities, trigonometric



equalities and inequalities, graphing of trigonometric functions, and polar coordinates.



Scores from this test should be used in conjunction with other mathematics test scores to



determine appropriate placement of the examinees (ACT, 2006).



Reliability



Because the COMPASS Tests are computer adaptive, each examinee is



administered a test consisting of different items each time the test is given. To make

69

conventional reliability formulas apply, the individual reliabilities of individual tests



must be averaged and used for comparative purposes. ACT uses standard error of



measurement (SEM) as a method of determining the reliability of this test instrument.



Using the SEM method for determining reliability, the standard length COMPASS tests



yields these reliability results: Reading, 0.78-0.79; Numerical/Prealgebra, 0.85; Algebra,



0.86; College Algebra, 0.85; Geometry, 0.88; and Trigonometry, 0.85. All of these



reliability figures are consistent with other standardized assessment instruments. The



mean score for the COMPASS Reading Test obtained in a study conducted by ACT for



two-year college students in the Fall of 2004 was 78.8 with a standard deviation score of



16.0. In the same study the COMPASS Algebra Test score mean was 31.2 with a



standard deviation of 18.4 (ACT, 2006).



Validity



Validity for the COMPASS Tests is measured by determining how well they act



as placement instruments for students entering postsecondary education. For the tests to



have high content validity in this instance, they must assess the knowledge and skills of



students that are important to success in college. Content validity for computer adaptive



tests is influenced by the representativeness of the items in the tests’ item pool and how



well they measure applicable skills. Content validity in this case is also influenced by



how representative each individual examinee’s computer adaptive test is of the skills



needed for college success. ACT suggests that the COMPASS Tests are valid



instruments for placement purposes because:



1. The COMPASS Tests measure the skills that students require for college



success.

70

2. Students who have the necessary skills for college success perform well on the



COMPASS Tests and students lacking those skills perform poorly on the



tests.



3. Higher levels of performance on the test are related to higher levels of college



success by students.



If these assertions are true, then there should be a positive relationship between



students’ COMPASS Test scores and their grades in entry-level college courses (ACT,



2006).



To measure the validity of its COMPASS Test, ACT uses placement validity



indices instead of the conventional practice of calculating correlation coefficients.



Coefficient correlations have traditionally been used to document the relationship



between test scores and course grades. However, this method has these disadvantages:



1. The correlation of placement tests with course grades can be easily



misinterpreted. In most instances, examinees scoring above an established



cut-off score will be placed in standard college course and those scoring



below that point will be placed in remedial or developmental courses. When



grades for the standard course in this instance are compared to test scores,



only those scores of the examinees placed in the course will be compared.



Examinees not meeting the minimum cut score will not be included in the



calculation. Without including all of the possible scores, accurate placement



by the test will restrict the number of examinees earning poor grades and will



decrease the correlation of course grades to test scores. This situation could

71

be misinterpreted by institutions as evidence of limited test validity when it



means just the opposite.



2. Correlations of grades and test scores assume that the distribution of grades



is normal and that the strength of the relationship is constant throughout the



range of scores. These assumptions are usually unsubstantiated and lead to



misinterpretation of correlation results.



The placement validity indices method practiced by ACT however allows for



curvilinear relationships and accounts for the differences in strength of the relationship



between individual test score and course grade. The placement validity index method



estimates the probability of success in the standard course by examinees placed in



remedial courses. This statistical method yields four estimated percentages:



1. The percentage of examinees that scored below the cutoff score and who



would have failed the standard course.



2. The percentage of examinees who scored below the cutoff score and who



would have succeeded in the standard course.



3. The percentage of examinees that scored at or above the cutoff and who



succeeded in the standard course.



4. The percentage of examinees that scored at or above the cutoff and who



failed the standard course.



Placement validation is accomplished by calculating the sum of the percentages of



examinees correctly placed or those found in 1 and 3 above. Using this method, the



COMPASS Test’s validity varies between 59-72% for examinees achieving a C or



better in their course work when the cutoff score for placement is defined as the

72

minimum score for which a examinees has a 50% chance of success. This means that



the placement validation index would have been 9-29% more accurate at placing



examinees than using an optimal cutoff score (ACT, 2006).



Clearly, the COMPASS Tests meet accepted standards for both reliability and



validity. In addition, ACT has adopted a system of test item generation and review that



continually confirms the tests’ reliability and validity. Its computer adaptive feature



adds a level of complexity to maintaining its reliability and validity but ACT has in



place the methods to address those complexities. The tests’ capability of generating



individual test forms for each examinee each time the tests is administered is an



advantage because examinees and institutions are not limited by time lapse requirements



for retesting. Its wide adoption by postsecondary institutions likewise, is a strong



indicator of its acceptance as an accurate placement instrument.



The information that follows presents and discusses in narrative and chart form,



the results obtained from analyzing the technical aspects, constructs measured, content,



and reliability of the COMPASS and GED Reading and Mathematics Tests. The first set



of tables present a side by side comparison of the similarities and differences between



technical aspects of the two instruments. The second set of tables describes the



similarities and differences between the content covered by the two instruments. This



comparison uses the COMPASS Tests as the baseline instruments because the



COMPASS Tests are designed to reliably determine the college readiness of students



(ACT, 2007) as opposed to the GED which corresponds to the material that a high



school graduate should know and have mastered (GEDTS, 2007). Accordingly, it is the



GED Tests that are being compared to the COMPASS Tests to determine if they can be

73

used for predicting college readiness. Comparison of what contents are measured by the



two tests were made by comparing similarities in descriptive language used by the



respective test technical manuals as well as a review of practice test items.



Comparison of GED and COMPASS Tests



Technical Similarities



The COMPASS and GED Tests are both commercially produced and possess



similar reliability and validity scores. Both tests use a multiple choice format with one



correct answer and four distracters. Both assess students’ abilities to use higher order



thinking skills such as comprehension, application, analysis, evaluation, and synthesis.



Technical Differences



The technical differences between the COMPASS and GED Tests are



considerable. The GED Reading and Mathematics Tests both have a set number of items



for each test while the COMPASS Tests, being computer-adaptive, may vary in item



number. The GED Tests have strict time limitations while the COMPASS Tests are



untimed. The score ranges of the two tests are also considerably different with



individual GED Tests having a top score of 800 while each of the COMPASS Tests



have maximum scores of 99 respectively. The COMPASS Mathematics Placement Test



also has the difference of being divided into five separate placement tests, Numerical



Skills/Prealgebra, Algebra, College Algebra, Geometry, and Trigonometry as compared



to a single mathematics test for the GED. The protocol for retesting also sets the two



tests apart. The GED again has a strict six month waiting period between individual test



administrations unless examinees successfully document completion of an accepted test



preparation course, while individuals may retest on the COMPASS without restriction.

74

Tables 1 and 2 follow and describe the technical similarities and differences between the



GED Reading and COMPASS Reading Tests and GED Mathematics and COMPASS



Mathematics Placement Tests respectively.

75

Table 1



Technical Comparison of GED Reading Test and COMPASS Reading Placement



Tests





Test Characteristic GED COMPASS



Reliability range .70 - .83 .78 - .88



Validity range .53 - .77 .59 - .72



Total test items 40 13 average



Time per test 65 minutes Untimed



Item configuration Multiple choice Multiple choice



Score range 0-800 0-99



Testing media Paper pencil, optical Computer adaptive,

mark score sheet Paper and pencil, optical

mark score sheet



Retest protocol Six month wait or Retest anytime

completion of a

preparation program







(ACT, 2006; GEDTS, 2007)

76

Table 2



Technical Comparison of GED Mathematics Test and COMPASS Mathematics



Placement Tests





Test Characteristic GED COMPASS



Reliability range .70 - .83 .85 - .88



Validity range .53 - .77 .59 - .72



Total items test 50 13.5 – 14.5



Time per test 90 minutes Untimed



Item configuration Multiple Choice, four Multiple Choice, four

distracters distracters



Score range 0-800 0-99



Testing media Paper and pencil, optical Computer adaptive, Paper

mark score sheet and pencil, optical mark

score sheet



Calculator use allowed Yes Yes



Retest protocol Six month wait or Retest anytime

completion of a

preparation program





(ACT, 2006; GEDTS, 2007)

77





From Tables 3 and 4, the two tests appear to have considerable agreement in the



domain of Reading regarding the content that they attempt to assess. Using the list of



reading skills identified in the COMPASS Technical Manuel as a baseline and then



identifying corresponding GED content language, it is clear that both tests attempt to



assess a complete range of reading skills. These skills range from the basic identification



and location of written information to the use of information in new and complex ways.



The language used to describe the content measured by the two test makers differs



markedly in some cases, but a review of the technical manuals, practice tests, and



preparation material, clearly indicates that they are seeking to measure a similar range of



content. Tables 3 and 4 follow.

78

Table 3



Content Comparison of COMPASS Reading Placement Test and GED Reading Test





COMPASS Reading Comprehension – Inferring Items



COMPASS GED Equivalent



Content Area Description Content area Description



Recognize stated main idea Summarize main idea



Locate explicitly stated information Restate or paraphrase information



Recognize sequential information Interpret patterns in text



Recognize cause and effect relationships Identify cause and effect

relationships



Recognize comparative relationships Compare and contrast



Recognize evidence supporting a claim Distinguish supporting statements



Recognize stated assumptions Understand consequences, make

inferences





(ACT, 2007; GEDTS, 2007)

79

Table 4



Content Comparison of COMPASS Reading Placement Test and GED Reading Test





COMPASS Reading Comprehension – Reasoning Items



COMPASS GED Equivalent



Content Area Description Content Area Description



Infer the main idea Make inferences



Show how details relate to the main idea Explain implications of text



Infer sequences Understand consequences



Infer cause and effect Identify cause and effect



Infer unstated assumptions Recognize unstated relationships



Draw conclusions from stated facts Explain implications of text



Make comparisons Compare and contrast



Make appropriate generalizations Integrate information; draw

conclusions



Recognize logical fallacies Distinguish supporting statements



Recognize stereotypes Recognize unstated assumptions



Recognize various points of view Interpret points of view



Recognize hypotheses, explanations, conclusions Integrate information



Judge relevance of and apply new information Make connections among parts of

text



Identify structure of an argument Distinguish unstated assumptions



Recognize relevant distinctions Compare and contrast



Supported and unsupported claims Distinguish unstated assumptions

80

COMPASS Reading Comprehension – Reasoning Items



COMPASS GED Equivalent



Content Area Description Content Area Description



Determine meaning from context Identify word usage

Apply information to new situations Integrate information from outside

the text





(ACT, 2007; GEDTS, 2007)

81

The content assessed by the COMPASS Numerical Skills/Prealgebra Placement Test



appears to be very closely related to the GED Mathematics Test with the exception of the item



described as Number Theory from the COMPASS Test Technical Manual’s list of content items.



No equivalent descriptor for Number Theory could be identified in any of the materials related to



the GED Mathematics Tests. Likewise, the COMPASS Algebra Placement Test and GED



Mathematics Test exhibit extensive overlap with only two items out of nineteen found as not



having an equivalent GED Test content descriptor. The COMPASS Geometry Placement Test



also has great similarity in content areas assessed with the GED Mathematics Tests and has



matching descriptors in 8 out of 10 content areas. Contrastingly, the COMPASS College Algebra



Placement Test and Trigonometry Placement Tests have little or no overlap with the GED



Mathematics Tests. The COMPASS College Algebra Placement Test had only two content



descriptors common to the GED Mathematics Test and the COMPASS Trigonometry Placement



Test had no content descriptors in common with the GED Mathematics Tests. Tables 5, 6, 7, 8,



and 9 follow.









Table 5

82

Content Comparison of COMPASS Numerical Skills/Prealgebra Placement Test and



GED Mathematics Test





COMPASS GED Equivalent



Content Area Description Content Area Description



Basic operations with integers Represent and use integers



Basic operations with fractions Represent and use Fractions



Basic operations with decimals Represent and use decimals



Exponents, square roots, scientific notation Represent and use exponents, and

scientific notation



Ratios and proportions Represent and use ratios and proportions



Percentages Represent and use percentages



Conversion of fractions and decimals Represent and use equivalent forms



Multiples and factors of integers Represent, analyze and apply integers



Absolute value of numbers Use algebraic expressions



Averages (medians, means modes) Apply measures of central tendency



Range Apply measures of central tendency



Order concepts (greater/less than) Equivalencies and order relationships



Estimation skills Use estimation to solve problems



Number theory No equivalent content



Counting problems and simple probability Make predictions based on probabilities





(ACT, 2007; GEDTS, 2007)

83

Table 6



Content Comparison of COMPASS Algebra Placement Test and GED Mathematics Test





COMPASS GED Equivalent



Content Area Description Content Area Description



Substituting values into algebraic Create and use algebraic

expressions expressions



Setting up equations for given Create and use algebraic

situations expressions



Basic operations with polynomials Create and use algebraic

expressions



Factoring of polynomials Create and use algebraic

expressions



Solving polynomial equations Create and use algebraic

by factoring expressions



Formula manipulation and field axioms Evaluate formulas



Linear equations, one variable Interpret, slope of a line, intersections



Rational expressions Create and use algebraic expressions



Exponents and radicals Analyze and use exponential functions



System of linear equations, two Solve equations using linear equations

variables



Quadratic formulas Solve quadratic equations



Absolute value equations and Factoring and inequalities

inequalities



Linear equations, two variables Solve equations using linear equations



Distance formulas in the plane No equivalent content

84



COMPASS GED Equivalent



Content Area Description Content Area Description



Graphing conics, circles, parabolas... Solve problems of length, area, perimeter



Graphing parallel lines Solves perpendicularity, parallelism...



Graphing relations in the plane Solve perpendicularity, parallelism...



Graphing equations and rational Graph generalized functional relationships

functions



Midpoint formulas No equivalent content





(ACT, 2007; GEDTS, 2007)

85

Table 7



Content Comparison of COMPASS College Algebra Placement Test and GED Mathematics Test





COMPASS GED Equivalent



Content Area Description Content Area Description



Functions Analyze and use functional relationships



Exponents Analyze and use functional relationships



Complex numbers No equivalent content



Arithmetic, geometric series No equivalent content

and sequences



Factorials No equivalent content



Matrices No equivalent content



Linear equations, 3 or more No equivalent content

variables



Algebraic logic and proofs No equivalent content



Roots of polynomials No equivalent content





(ACT, 2007; GEDTS, 2007)

86

Table 8



Content Comparison of COMPASS Geometry Mathematics Placement Test and GED



Mathematics Test





COMPASS GED Equivalent



Content Area Description Content Area Description



Triangles, Pythagorean Theorem Use Pythagorean Theorem to solve problems



Circles Solve problems of length, area, and perimeter



Angles No equivalent content



Rectangles Solve problems of length, area, perimeter



Three–dimensional concepts Analyze geometric figures



Hybrid shapes Analyze geometric figures



Trapezoids Analyze geometric figures



Parallelograms Analyze geometric figures



Geometric logic and proofs No equivalent content





(ACT, 2007; GEDTS, 2007)

87

Table 9



Content Comparison of COMPASS Trigonometry Mathematics Placement Test and GED



Mathematics Test





COMPASS GED Equivalent



Content Area Description Content Area Description



Trigonometric functions and identities No equivalent content



Right triangle trigonometry No equivalent content



Trigonometric equalities and inequalities No equivalent content



Graphs of trigonometric functions No equivalent content



Special angles No equivalent content



Polar coordinates No equivalent content





(ACT, 2007; GEDTS, 2007)

88

Summary of Content and Reliability Analysis



It is clear from the technical materials and practice test materials examined that



the GED and COMPASS Tests possess a high degree of similarity in reliability for



Reading. Additionally the content examined indicates that both tests appear to be



measuring higher order thinking skills such as inference, analysis, comprehension,



evaluation, and synthesis as well as common reading application skills. However, even



with the differences in formats between the GED Mathematics Tests and the five levels



of COMPASS Mathematics Tests it is clear from the content of those tests that the GED



Mathematics Test is more similar to the COMPASS Numerical Skills /Prealgebra Test,



COMPASS Algebra, and the COMPASS Geometry Test than it is to the COMPASS



College Algebra, and Trigonometry Tests.



The confirmation of the two Reading tests’ similarities in reliability and content



suggested that a linkage study of those tests would yield meaningful results. Likewise, it



appeared that a linking study of the GED Mathematics Tests with the COMPASS



Numerical Skills/Prealgebra, Algebra and Geometry Tests would also be likely to yield



meaningful results because of those tests’ similarities in reliability and in the content they



assess. However, the differences between the content descriptors of the GED Mathematics



Tests and the COMPASS College Algebra and Trigonometry Tests were great and had



almost no overlap. Therefore, they were not included in the analysis designed to link the



scores of the two tests because such inclusion was not likely to produce meaningful



results. Although the COMPASS Numerical Skills/Prealgebra Test had a high degree of



similarity to the GED Mathematics Test, it was not included for regression analysis



because no score on that test, not even a perfect score, is an indicator of college readiness

89

.Likewise, The COMPASS Geometry Test was not be included in the linear regression



analysis because the COMPASS Algebra Test was used as the primary instrument to



determine college readiness (HCCS, 2004). Given these findings, a linear regression



analysis was conducted to determine the degree to which GED Reading Test scores can



predict COMPASS Reading Placement Test scores. A similar analysis was conducted



between GED Mathematics Test scores and COMPASS Algebra Test scores (HCCS,



2004). The following section discusses those findings.



Quantitative Data Analysis Findings



Descriptive Statistics



The test scores included in this study were from GED and COMPASS Tests taken



by students enrolled in semester credit courses at Houston Community College during the



2006 calendar year and who had successfully completed the GED Tests between the



dates of January 1, 2002 and December 31, 2006. These parameters were chosen for



selection of the study’s data sets for these reasons: (1) On January 1, 2002, the version of



the GED Tests currently in use was put into service. (2) Only the scores from GED



completers who were enrolled in semester credit hour courses were included in the data



set because other course offerings at the college in which GED completers might have



enrolled may not have required a COMPASS test for admission purposes. (3) Between



January 1, 2006 and June 30, 2006, the State of Texas conducted a pilot project at



Houston Community College that required potential GED examinees to pass a



prequalification exam before attempting the official GED Tests.



The parameters described yielded ninety-one total records that were used in the



study. By gender, 34.1% of the study subjects were male and 65.9% of the study subjects

90

were female. The ethnic make up of the study group was 12.1% White; 47.3% Hispanic;



31.9% African American; Asian 5.5% and 3.3% had no ethnic designation. The study



subjects’ ages ranged from 17 to 50 years with a mean of 25.1. On average, subjects left



school sometime during the tenth grade, with the range for school exit being as early as



sixth grade and as late as the twelfth grade.



Data collected and analyzed for the descriptive analytic portions of the study



included COMPASS Test scores for Reading, Numerical Skills/Prealgebra and Algebra.



GED Test score data collected for the study included GED Reading Test scores, GED



Mathematics Test Scores, and GED Test score averages. Also included in the data



gathered were age, ethnicity, grade at time of school exit, percentile rank for GED



Reading Test score, and percentile rank for GED Mathematics Test score. Table 10 along



with Figures 1, 2, 3, 4, 5 and 6 describe those data in more detail.

91

Table 10



Study Subject Data Descriptive Statistics





Data Type N Minimum Maximum Mean Std. Dev.



COMPASS Reading Test Score 89 44 99 81.55 12.563



COMPASS Numerical/Prealgebra 86 17 96 41.83 20.622

Test Score



COMPASS Algebra Test Score 87 15 89 24.71 14.770



GED Reading Test Score 90 360 800 557.67 94.072



GED Mathematics Test Score 91 240 800 483.01 93.939



GED Test Battery Score Average 91 360 744 518.75 65.524



GED Reading Percentile Rank 90 8 99 65.08 24.466



GED Mathematics Percentile Rank 91 1 99 41.82 24.595



Subject Age at GED Test 91 17 50 25.21 8.697



Grade at School Exit 88 6 12 10.18 1.866

92





Ethnicity

12.5% White

48.9% Hispanic

5.7

12.5

32.9% African American

5.7% Asian









33









48.9









Figure 1. Ethnic make up of the study group.







Gender

34.1% Male

65.9% Female









34.1









65.9









Figure 2. Gender make up of the total study group.

93





12.5









10.0









Count

7.5

14

.0







11

.0

5.0

9.

0



7. 7.

0 0

6.

0

2.5



3. 3. 3. 3.

0 0 0 0

2. 2. 2. 2. 2. 2. 2. 2. 2.

0 0 0 0 0 0 0 0 0

1. 1. 1. 1. 1. 1. 1.

0 0 0 0 0 0 0

0.0

17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 37 38 44 45 46 47 48 50



AGE







Figure 3. Age at which the study subjects completed the GED.





30









25









20

Frequency









15 29

27









10









11

5 10







5 Mean =10.07

4 Std. Dev. =1.38

2 N =88

0

6 8 10 12



EXITGRAD







Figure 4. Grades at which the study subjects withdrew from school.

94





100









90









80

READ



D









70









60









50









40





0 20 40 60 80 100



READPCT





Figure 5. COMPASS Reading Score and GED Reading Score Percentile Rank.







Figure 5 illustrates the relationship between subjects’ COMPASS Reading Test score and



their percentile rank among GED Reading Test completers nationally. The horizontal line



represents the college readiness minimum COMPASS Reading Test score of 81. The



vertical line represents the boundary for the first quartile of GED Reading Test



completers. In this sample, 51 out of 89 subjects with COMPASS Reading Test scores



are indicated as college ready. The GED Testing Service norms GED Test scores to



indicate how GED completers would compare on a class rank basis to high school



graduates. For instance, a GED Test score of 500 on any of the five GED tests indicates a



class rank at the fiftieth percentile nationally. This example indicates that GED

95



completers may achieve college readiness COMPASS Reading scores even if they are



well below the top quarter of their predicted high school class rank. In contrast, some



GED completers who are indicated as being in the top quartile of high school graduates



nationally, did not achieve college ready COMPASS Reading Test



scores.









80









60

ALG









40









20









0 20 40 60 80 100



MATHPCT



Figure 6. COMPASS Algebra Score and GED Mathematics Percentile Rank.

96

Figure 6 illustrates the relationship between subjects’ COMPASS Algebra Test



score and their percentile rank among GED Mathematics Test completers nationally. The



horizontal line represents the college readiness minimum COMPASS Algebra Test score



of 71. The vertical line represents the boundary for the first quartile of GED Mathematics



Test completers. For this group of subjects, only one individual is indicated as college



ready. This example indicates that even GED completers who rank in the top quartile of



GED completers nationally are unlikely achieve college readiness in Mathematics.



Regression Analysis Findings



Frequency testing on the data set used in the study identified no skewness in the



scores and confirmed that the study’s data could be considered normally distributed and



any resulting relationships as linear. No data transformations were necessary. A



hierarchical multiple regression analysis was conducted to determine the degree to which



GED Reading Test scores [GEDREAD] and GED Mathematics Test scores



[GEDMATH] predicted COMPASS Reading and Algebra Test scores, respectively.



Additional independent variables (age of the student when they obtained their GED



[AGE], student gender [Gender], student ethnicity [Ethnicity], the grade at which they



dropped out of school [EXITGRAD], and whether or not they had attended an adult



education class at Houston Community College [AECLASS]) were included as



independent variables.



Four separate block models were run for each subject area. The regression



analysis using ALG as the dependent variable showed that in combination, Age, Ethnicity



and Gender accounted for only 9.8% of the model’s ability to predict COMPASS Algebra



scores, and in combination with other demographic independent variables accounted for

97



40.5% of the variance. The first, second and third models in Table 11 indicate



significance for the independent variable Age at the p<.05 level. The full model that



included GEDMATH had an R2 =.405 (F (6, 77) = 8.732 and a significance at the



p<.001), indicating the strong predictive value-added on COMPASS Algebra scores.



Tables 11 and 12 present the unstandardized coefficients and standard errors of those



models.



Table 11



B Coefficients for ALG as the Dependent Variable



Model 1 Model 2 Model 3 Model 4

(sd) (sd) (sd) (sd)



Age -.415** -.414** -.423** .009

(.180) (.181) (.188) (.169)



Gender -4.611 -4.350 -.4.308 -.672

(3.475) (3.507) (3.536) (2.958)



Ethnicity -.426 -.544 -.586 1.201

(2.229) (2.244) (2.268) (1.882)



EXITGRAD .789 .849 .315

(1.167) (1.203) (.991)



AECLASS .684 -1.119

(3.592) (2.960)



GEDMATH .099*

(.016)



* significant at p<.001

** significant at p<.05

98



Table 12



B Coefficients for READ as the Dependent Variable



Model 1 Model 2 Model 3 Model 4

(sd) (sd) (sd) (sd)



Age -.233 -.232 -.228 -.177

(.159) (.158) (.163) (.147)



Gender -4.860 -4.638 -.4.644 -2.149

(2.941) (2.938) (2.957) (2.703)



Ethnicity -1.374 -.1.535 -1.522 -.753

(1.882) (1.881) (1.897) (1.706)



EXITGRAD 1.181 1.163 .748

(.969) (.990) (.891)



AECLASS -.315 -.069

(3.006) (2.691)



GEDREAD .061*

(.014)



* significant at p<.001



In the second set of models that used COMPASS READ as the dependent



variable, the regression analysis showed Age, Ethnicity, and Gender accounted for only



8.4% of the model’s ability to predict COMPASS READ scores, and in combination with



other demographic independent variables accounted for 28.9% the variance. The full



model that includes GEDREAD has an R2 =.289 (F (6, 78) = 5.277, p<.001) indicating



that GEDREAD scores have strong predictive value for COMPASS READ scores even



after controlling for other contributing variables.

99



Equipercentile Scaling Findings



The results of the regression analysis suggest there is strong evidence that the



GED Mathematics and Reading Tests and the COMPASS Algebra and Reading Tests are



similar to the degree that their scores may be meaningfully linked. If the scores for the



GED Tests under consideration can be used in certain instances to predict COMPASS



Test scores then they are useful also for predicting college readiness. To make the



linkages between the GED and COMPASS Test scores more apparent for predictive



purposes, a series of concordances tables have been developed using the equipercentile



scaling method. In instances where the percentile ranks between the two tests’ scores



were not exact, those scores were concorded to within two percentile points of each other.



In instances where there were no corresponding scores that met that criterion for



concordance, no score was assigned. Because of the study’s relatively small sample size,



this situation sometimes resulted in a GED score having no COMPASS equivalent and at



other times with COMPASS scores having no GED equivalent. Resultantly, Table 13



(See Appendix A) shows that a GED score of at least 540 would necessary to predict that



a student is college ready in the domain of Reading, while Table 14 (See Appendix A)



indicates that a student would have to obtain a nearly perfect score of 790 in the domain



of Mathematics, to be considered college ready.



Summary of Quantitative Findings



This chapter presented the findings of the content analysis, linear regression



analysis, and equipercentile scaling procedure used to determine the relationships



between GED and COMPASS Test scores. The results of the content analysis confirmed



that while there are differences between the GED and COMPASS Tests under

100



consideration, they are similar enough in the content they measure and in reliability for



their test scores to have meaningful test score linkages. The regression analyses of those



linkages suggests that some were significant to the degree that made GED Reading Test



scores predictive of COMPASS Reading Tests scores and GED Mathematics Test scores



predictive of COMPASS Algebra Test scores. The equipercentile scaling procedure



conducted on the scores collected for the study suggest that a score of 540 on the GED



Reading Test concords to the college ready score of 81 on the COMPASS Reading Test



while a score of 790 on the GED Mathematics Test concords to a college ready score of



71 on the COMPASS Algebra Test. The final chapter of the study, will discuss the



study’s results in light of the hypotheses set forth at the beginning of this study and will



conclude with a discussion of its findings’ implications for practice and future research.

CHAPTER FIVE



CONCLUSIONS AND SUMMARY



To determine if the GED Test scores can be used as indicators for college



readiness, this study posed five research questions relative to the GED and COMPASS



Tests. Those five research questions were:



1. To what degree are the tests measuring the same content?



2. To what degree are the tests similarly reliable?



3. To what degree are the tests symmetrical?



4. To what degree are the tests’ scores linkable?



5. To what degree are any test score linkages population invariant?



To answer the first two research questions, a content analysis of materials



associated with the COMPASS Reading and Mathematics Placement Tests and the GED



Reading and Mathematics Tests was conducted. This content analysis identified



similarities and differences between the content measured by the tests and determined



how similar the tests were in reliability. To answer the last three research questions, a



regression analysis of the applicable GED and COMPASS Test scores was conducted.



The regression analysis provided data regarding the strength of any linkages between the



scores of the respective tests. Additionally, an equipercentile scaling procedure was



conducted that resulted in concordance tables comparing GED Reading and COMPASS



Reading scores and comparing GED Mathematics Test scores to COMPASS Algebra



Test scores. The concordance tables were developed after regression analyses results



suggested that the scores of the GED and COMPASS Tests under consideration were

102



related to a significant degree. The discussion that follows describes the results of those



analyses and procedures.



Discussion



To What Degree Are The Tests Measuring The Same Content?



The first hypothesis explored by this study suggesting that the content measured



by the GED Reading and Mathematics Placement Tests are similar to the content



measured by the COMPASS Reading and Mathematics Placement Tests is partially



confirmed. Results of the content analysis clearly indicate that the Reading Tests for both



COMPASS and GED are assessing similar content. Both reading tests assess higher order



thinking skills like comprehension, analysis, synthesis, and evaluation. The content



analysis of the COMPASS Mathematics Placement Tests was complicated by the fact



that there are five separate mathematics tests assessing five separate areas of mathematics



content. The GED Mathematics Test on the other hand, is a single instrument that



assesses a wide range of mathematics content. The COMPASS Numerical



Skills/Prealgebra, Algebra, and Geometry Placement Tests share much in common with



the GED Mathematics Test and are clearly measuring similar content. In the case of the



COMPASS Numerical Skills/Prealgebra Placement Tests in particular, there is complete



overlap of its content descriptors and those found in the analysis of the GED Mathematics



Tests’ materials. This is not the case with all of the COMPASS Mathematics Placement



Tests however. Specifically, the COMPASS College Algebra Placement Test and



Trigonometry Placement Test have few if any common content descriptors relative to the



GED Mathematics Tests. This lack of similarity reduces the likelihood that the scores of



those particular tests can be meaningfully linked.

103



To What Degree Are The Tests Similarly Reliable?



The second hypothesis explored by this study suggesting that the reliability scores



of the GED Reading and Mathematics Tests are similar to the reliability scores of the



COMPASS Reading and Mathematics Placement Tests is confirmed. With reliability



scores ranging from .70 - .83 for the GED Tests and .78 - .88 for the COMPASS tests, the



reliability scores of the COMPASS Tests and the GED Tests are clearly similar. While



the two tests are considerably different in many of their technical aspects, these



differences are overshadowed by the strong similarities of the tests reliability and validity



scores, indicating that their respective test scores can be meaningfully linked.



To What Degree Are The Tests Symmetrical?



The third hypothesis explored by this study suggesting that the scores of the GED



Reading and Mathematics Tests are symmetric with the scores of the COMPASS



Reading and Mathematics Placement Tests is inconclusive. The data obtained and



analyzed for this study was not sufficient to determine the degree to which the GED and



COMPASS Tests under consideration could be considered symmetrical. As was



discussed in the study’s results section, all of the students in this study passed the GED



Tests before attempting the COMPASS Tests. No data were available from students who



had first taken the COMPASS Tests and subsequently attempted and passed the



corresponding GED Tests. Without such data, no analysis can be conducted to determine



the degree to which GED Test and COMPASS Tests are symmetrical.



To What Degree Are The Tests’ Scores Linkable?



The fourth hypothesis explored by this study suggesting that the scores of the



GED Reading and Mathematics Tests can be meaningful linked to the scores of the

104



COMPASS Reading and Mathematics Placement Tests is partially confirmed. The two



tests meet the expectation for linkage studies in regard to measuring similar content and



being similar in reliability for the domain of Reading and for the COMPASS Algebra



Test which the primary instrument used by HCC to determine college readiness in



mathematics. The results of the regression analyses comparing GED and COMPASS Test



scores indicate significant relationships at the p<.001 level between their scores in



Reading and for GED Mathematics scores relative to COMPASS Algebra scores.



To What Degree Are Any Test Score Linkages Population Invariant?



The fifth and final hypothesis explored by this study suggesting that any linkages



of GED Reading and Mathematics Tests scores will be population invariant is



inconclusive. The sample size used in this study was too small to conduct reliable



regression analyses by subgroup. However, because ethnicity and gender were not



significant contributors to the variance found in COMPASS Test scores, it is likely that



linkages between the two tests are population invariant.



Implications for Practice



The results of the equipercentile scaling procedure suggest that the college



readiness score established for the COMPASS Reading Test of 81 concords to a score of



at least 540 on the GED Reading Test. The minimum passing score for the GED Reading



Test is 410. About half of the subjects in this study achieved COMPASS college



readiness scores in Reading. No subject in the study scored exactly 81 on the COMPASS



Reading Test. However, a GED Reading Scores of 530 concorded to a COMPASS



Reading score of 80, while a GED Reading score of 540 concorded to a COMPASS



Reading Score of 83. GED Reading Test scores are accumulated in ten point increments

105



and therefore it is logical to conclude that scoring 540 of the GED Reading test will likely



mean that a student is college ready in the domain of Reading.



In the domain of Mathematics however, the concordance of scores suggests that a



GED Mathematics score of at least 790 is necessary to achieve college readiness. Among



the subjects for this study, only one out of 91 subjects scored well enough on the



COMPASS Algebra Test to be considered college ready. A score of 780 on the GED



Mathematics Test concorded to a score of 69 on the COMPASS Algebra Test. A perfect



score of 800 on the GED Mathematics Test concorded to a score of 89 on the COMPASS



Algebra Test for this sample.



In combination, the results of the content analysis conducted as part of this study



clearly showed that the GED and COMPASS Tests possess great similarities in content



for the domains of Reading and Mathematics. The respective tests were also found to be



likewise similar in reliability. The significant results of the regression analysis of the



scores of the tests further strengthen the assertion that their scores can be meaningfully



linked. These results confirm the work by researchers who suggest that the scores of two



different tests can be meaningfully linked if they are similar enough in content and



reliability (Dorans & Holland, 2000; Kolen & Brennan, 2004). The regression analysis



results of this study also suggest the scores of the selected GED Tests appear to be useful



for predicting college readiness and confirm earlier studies that explore the use of other



standardized tests for that purpose (DeBerard, Spelmans, & Julka, 2004; Garavalia &



Gredler, 2002; Mulvenon, Stegman, & Thorn, 1999; Naumann, Bandalos, & Gutkin,



2003; Popham, 2006).

106



A concordance of scores between the GED and COMPASS is important to



institutions like Houston Community College for two reasons. First, such concordance



tables are useful to teachers who prepare students to attempt and pass the GED Tests.



Having a more research-based insight into the level of proficiency that GED completers



must acquire to be college ready informs teachers regarding how best to structure their



curricula based on the educational and career goals of their students. Not every GED



completer intends to enroll in postsecondary education but those that do clearly need to



engage a more extended mathematics curriculum. Second, concordance tables like these



are useful to the college staff persons responsible for student advisement. The additional



information that GED scores can provide to advisement staff persons can assist them to



communicate with GED completers regarding the gap that exists between college



readiness and the minimum proficiencies required to pass the GED Tests. By providing a



link between research and practice, these concordance tables add value to the research



conducted in this study and make its results more likely to benefit adult learners



preparing to complete the GED Tests and subsequently enroll in Houston Community



College or other post secondary institutions.



The results of the regression analyses conducted for this study indicate that scores



from the GED Reading and Mathematics Tests are significantly related to the scores of



the COMPASS Reading and COMPASS Algebra Tests respectively. This means that



with some caution, GED Test scores may be useful to predict college readiness for GED



completers.



The results of the study’s equipercentile scaling analysis suggests that a score of



540 on the GED Reading Test concords to a COMPASS Reading Test score of 81, the

107



minimum score necessary for students to be considered college ready in the domain of



Reading. There were no subjects in the sample with a COMPASS Algebra score of 71,



the minimum college readiness score in mathematics. However, a GED Mathematics Test



score of 780 concords to a COMPASS Algebra Test score of 69 and a perfect score of



800 on the GED Mathematics Test concords to a COMPASS Algebra Test Score of 89 in



this sample. This relationship suggests that to achieve the minimum college readiness



score of 71, a student should score at least a 790 on the COMPASS Algebra Test. Given



that scores near the maximum possible on the GED Mathematics Test concord to the



minimum college readiness score on the COMPASS Algebra Test, it is advisable that the



curriculum used to prepare students for the GED Mathematics Test be expanded to



include more Algebra-specific content when those students indicate enrollment in college



as an educational goal.



Limitations



Practitioners and researchers are urged to exercise caution when applying the



results of this study due to the small size and uniqueness of the sample used. The small



sample size prevented the determination of the research question related to population



invariance because subpopulation numbers were too small to yield meaningful results. In



addition, the research question relative to test symmetry could not be answered because



all of the subjects in this sample had taken the GED Tests prior to attempting the



COMPASS Tests and no reciprocal data were available from subjects who had taken the



COMPASS Tests before taking the GED Tests. The study also does not analyze the effect



of any student personal characteristics of GED completers that might have an influence



on their college readiness. Studies of high school graduates clearly indicate that personal

108



attributes account for a far greater portion of the variance of college readiness measures



than do standardized test scores like the ACT and SAT (DeBerard, Spelmans, & Julka,



2004; Garavalia & Gredler, 2002; Mulvenon, Stegman, & Thorn, 1999; Naumann,



Bandalos, & Gutkin, 2003; Popham, 2006). It is likely then that the personal attributes of



GED completers could have a large influence on their college readiness as well but that



possibility was not tested in this study.



Suggestions for Future Research



Given the information cited in the review of the literature conducted for this



study, it is clear that there is a dearth of research that looks at predicting the college



readiness and college success for GED completers. To address one of the main



limitations of this study, a study using a larger sample should be conducted to determine



the extent to which the score linkages between the two tests are population invariant



across the subgroups of age, ethnicity, gender and ethnicity and gender in combination. A



second limitation of this study prevented it from determining the extent to which the two



tests are symmetrical. To determine if the tests possess a meaningful degree of symmetry,



a study should be conducted that looks at the relationship of GED and COMPASS scores



for students who first take the COMPASS Tests and subsequently attempt the GED Tests.



Further studies might also compare the college readiness of GED completers who



attended preparation courses to those completers who took the GED Tests without the



benefit of any test preparation. Finally, studies that look at the effect of personal traits as



well as academic preparation on college readiness for GED completers should be



conducted to determine if those variables affect college readiness for GED completers in



ways similar to high school graduates.

109



Conclusion



To meet the needs of employers in Texas for a trained workforce, more of its



traditional college age population and more nontraditional students must be enrolled,



retained, and graduated from its four-year universities and community colleges (Arnone,



2003a). GED completers are recognized as an important source of potential college



students but often are underprepared to be successful in college level coursework



(CALEC, 1993b). This study is significant and supports the “Texas Success Initiative” of



the Texas Legislature and “Closing the Gaps” initiative of the Texas Higher Education



Coordinating Board. The study also benefits adult learners because by providing



research-based information to community colleges and adult education providers with



information regarding the level of achievement required for GED recipients to be



considered college ready. By accurately advising adult learners preparing for the GED



regarding the level of achievement in Reading and Mathematics required to be college



ready, GED preparation providers and adult education providers can assist those students



to avoid the expense in time and money of placement into developmental studies



coursework and increase the likelihood that they complete an occupational certificate or



degree.

BIBLIOGRAPHY



ACE. (2007a). GED information for education and testing professionals. American



Council on Education, 2007. Retrieved February 5, 2007 from



http://www.acenet.edu/AM/Template.cfm?Section=Professionals&Template=/Ta



ggedPage/TaggedPageDisplay.cfm&TPLID=58&ContentID=7788.



ACE. (2007b). What's next? From GED to college graduate. American Council on



Education. Retrieved July, 4 2007 from



http://www.acenet.edu/AM/Template.cfm?Section=GEDTS&Template=/TaggedP



age/TaggedPageDisplay.cfm&TPLID=58&ContentID=17971.



ACT. (2004a). COMPASS/ESL. Sample test questions- A guide for students and parents.



Mathematics. Numerical skills, pre-algebra, algebra, Retrieved January 9, 2008



from http://www.austincc.edu/tutor/students/assessmentprep.php.



ACT. (2004b). COMPASS/ESL. Sample test questions- A guide for students and parents.



Reading, Retrieved January 9, 2008 from



http://www.austincc.edu/tutor/students/assessmentprep.php.



ACT. (2006). COMPASS/ESL: A comprehensive computerized placement and diagnostic



assessment and support system. Reference manual. Iowa City, IA: ACT, Inc.



ACT. (2007). ACT's COMPASS system, 2007. Retrieved July 15, 2007 from



http://www.act.org/compass/placement.html.



Arnone, M. (2003a). A Texas-size challenge. The Chronicle of Higher Education, 50(14),



A13-15. Retrieved April 14, 2007 from The WilsonWeb.

111



Arnone, M. (2003b). Texas falls behind in plan to enroll more minority students. The



Chronicle of Higher Education, 49(19), A23. Retrieved April 14, 2007 from The



WilsonWeb.



ASSCB. (2006). Spellings report spells out future for higher ed in U.S. Association to



Advance Collegiate Schools of Business, November/December.



Baycich, D. (2003). GED grads in college: Building awareness. Adult Learning, 14(3), 5-



7. Retrieved April 7, 2007 from The WilsonWeb.



Burke, J. C. (2005). Closing the accountability gap for public universities: Putting



academic departments in the performance loop. Planning for Higher Education,



34(1), 19-28. Retrieved April 3, 2007 from The WilsonWeb.



Burke, J. C., & Minassians, H. P. (2004). Implications of state performance indicators for



community college assessment. New Directions for Community Colleges (126),



53-64. Retrieved October 6, 2007 from The WilsonWeb.



CALEC. (1993a). GED uses public TV to assist adult learners with "GED--get it!"



Educational Record, 74(Fall), 57. Retrieved June 26, 2007 from Center for Adult



Learning and Educational Credentials.



CALEC. (1993b). Record number of GED graduates plan on further education.



Educational Record, 74, 61. Retrieved June 15, 2007 from Center for Adult



Learning and Educational Credentials.



CALEC. (1995). The difference age makes: GED profiles. Educational Record, 76, 60.



Retrieved April 7, 2007 from Center for Adult Learning and Educational



Credentials.

112



Carey, K. (2007). Truth without action: The myth of higher-education accountability.



Change, 39(5), 24-29. Retrieved July 15, 2007 from The WilsonWeb.



Carlan, P. E., & Byxbe, F. R. (2000). Community colleges under the microscope: an



analysis of performance predictors for native and transfer students. Community



College Review, 28(2), 27-42. Retrieved June 2, 2007 from The WilsonWeb.



Conley, D. T. (2007). Toward a more comprehensive conception of college readiness.



Eugene, OR: Education Policy Improvement Center.



DeBerard, M. S., Spelmans, G. I., & Julka, D. L. (2004). Predictors of academic



achievement and retention among college freshmen: A longitudinal study. College



Student Journal, 38(1), 66-80. Retrieved June 2, 2007 from The WilsonWeb.



Dorans, N. J., & Holland, P. W. (2000). Population invariance and the equatability of



tests: Basic theory and the linear case. Journal of Educational Measurement,



37(4), 281-306. Retrieved September 3, 2007 from The JSTOR Archive.



Dozier, S. B. (2001). Undocumented and documented international students: a



comparative study of their academic performance. Community College Review,



29(2), 43-53. Retrieved June 2, 2007 from The WilsonWeb.



Fisher, M. E., & Sandiford, J. R. (2000). A comparison of academic performance of



students with general education development credentials and high school



graduates and a selected community college. Journal of Applied Research at the



Community College, 8(1), 5-13. Retrieved June 2, 2007 from The WilsonWeb.



Garavalia, L. S., & Gredler, M. E. (2002). Prior achievement, aptitude, and use of



learning strategies as predictors of college student achievement. College Student



Journal, 36(4), 616-625. Retrieved June 2, 2007 from The WilsonWeb.

113



GEDTS. (2003a). Tests of General Educational Development. Official Practice Test. Test



Form PE, Mathematics, Part I. Washington, DC: GED Testing Service. American



Council on Education.



GEDTS. (2003b). Tests of General Educational Development. Official Practice Test. Test



Form PE, Mathematics, Part II. Washington, DC: GED Testing Service.



American Council on Education.



GEDTS. (2003c). Tests of General Educational Development. Official Practice Test. Test



Form PE. Language Arts, Reading. Washington, DC: GED Testing Service.



American Council on Education.



GEDTS. (2007). The 2002 series tests of General Educational Development: Technical



manual. (In press). Washington, DC: American Council on Education.



Golden, S. (2003). Self efficacy: How does it influence academic success? Adult



Learning, 14(3), 14-16. Retrieved June 2, 2007 from The WilsonWeb.



Hamilton, J. (1998). First time students entering a two-year pubic community college



with a GED. Fall 1991 to fall 1996. Gainesville. Georgia: Gainesville College.



HCCS. (2002). GED 2002 transitions: A guide for chief examiners: Houston Community



College System.



HCCS. (2004). Houston Community College System Texas success initiative plan, 2007



from http://www.hccs.edu/dev/plan.htm.



HCCS. (2006). Houston Community College System GED test results, 2007 from



http://www.hccs.edu/students/GED/results.htm.



Hoyt, J. E. (1999). Remedial education and student attrition. Community College Review,



27(2), 51-72. Retrieved October 7, 2007 from The WilsonWeb.

114



IES. (2007). State education data profiles. National Center for Education Statistics.



Retrieved October 18, 2008, from http://nces.ed.gov/programs/stateprofiles/.



Kolen, M. J., & Brennan, R. L. (2004). Test equating, scaling, and linking: methods and



practice (2nd ed.). New York: Springer Science+Business Media, Inc.



Lamkin, M. D. (2004). To achieve the dream, first look at the facts. Change, 36(6), 12-



15. Retrieved February 12, 2007 from The WilsonWeb.



Lau, L. K. (2003). Institutional factors affecting student retention. Education, 124(1),



126-136. Retrieved June 2, 2007 from The WilsonWeb.



Laverge, G. M., & Walker, B. (2006). Developing a concordance between the ACT



assessment and the SAT I: reasoning test for the University of Texas at Austin,



2008 from http://www.utexas.edu/student/admissions/reseacrh/ACT-



SATconcordance.html.



LBB. (2007). Texas Legislature Conference Committee Report on HB1, III-21 from



http://www.lbb.state.tx.us/Bill_80/7_Conference/80-7_Conference_0507.pdf.



Linn, R. L. (1993). Linking results of distinct assessments. Applied Measurement in



Education, 6(1), 83-102.



McLendon, M. K., Hearn, J. C., & Deaton, R. (2006). Called to account: Analyzing the



origins and spread of state performance-accountability policies for higher



education. Educational Evaluation & Policy Analysis, 28(1), 1-24. Retrieved



October 7, 2007 from The WilsonWeb.

115



Miglietti, C. L., & Strange, C. (1998). Learning styles, classroom environment



preferences, teaching styles, and remedial course outcomes for underprepared



adults at a two-year college. Community College Review, 26(1), 1-19. Retrieved



October 7, 2007 from The WilsonWeb.



Mislevy, R. J. (1992). Linking educational assessments: Concepts, issues, methods and



prospects. Princeton, N.J.: Education Testing Service.



Mulvenon, S. W., Stegman, C. E., & Thorn, A. (1999). Selection for college admission:



refining tradition models. Journal of College Admission (162), 20-27. Retrieved



June 2, 2007 from The WilsonWeb.



Naumann, W. C., Bandalos, D., & Gutkin, T. B. (2003). Identifying variables that predict



college success for first-generation college students. Journal of College



Admission (181), 4-9. Retrieved June 2, 2007 from The WilsonWeb.



NCWE. (2004). Breaking through: Helping low-skilled adults enter and succeed in



college and the workplace. National Council for Workforce Education, 2007 from



http://www.ncwe.org/documents/report_2004_ncweJff_breakingThrough.pdf.



Northcutt, E. (Ed.). (2002). Complete GED preparation. Austin, Texas: Steck-Vaughn



Company.



O'Neill, S. P. (1995). A study of the academic progress of high risk GED graduates



enrolled in an urban community college. Adult Basic Education, 5(1), 28-36.



Osborne, J. W. (1997). Identification with academics and academic success among



community college students. Community College Review, 25, 59-67. Retrieved



June 2, 2007 from The WilsonWeb.

116



Pathey-Chavez, G. G., Dillon P.H. & Thomas-Spiegel, J. (2005). How far do they get?



Tracking students with different literacies through community college



remediation. Teaching English in the Two Year College, 32(3), 261-278.



Retrieved June 2, 2007 from The WilsonWeb.



Popham, W. J. (2006). Branded by a test. Educational Leadership, 63(7), 86-87.



Retrieved June 2, 2007 from The WilsonWeb.



Pusser, B., Breneman, D. W., Gansneder, B. M., Kohl, K. J., Levin, J. S., Milam, J. H., et



al. (2007). Returning to learning: Adult's success in college is key to America's



future, 2007 from



http://www.luminafoundation.org/publications/ReturntolearningApril2007.pdf.



Ridgell, S. D., & Lounsbury, J. W. (2004). Predicting academic success: General



intelligence, "big five" personality traits, and work drive. College Student



Journal, 38(4), 607-618. Retrieved June 2, 2007 from The WilsonWeb.



Rose, M. C. (1999). Using ACT and GED as indicators of success for postsecondary



secondary students enrolled with GED certificates. Research and Teaching in



Developmental Studies, 15(2), 55-62.



Rosenthal, B. S., & Wilson, W. C. (2003). Impact of exposure to community violence



and psychological symptoms on college performance among students of color.



Adolescence, 38, 239-249. Retrieved June 2, 2007 from The WilsonWeb.



Roth, J., Crans, G. G., & Carter, R. L. (2000). Effect of high school course-taking and



grades on passing a college placement test. The High School Journal, 84(2), 72-



87. Retrieved June 2, 2007 from The WilsonWeb.

117



Schmid, C., & Abell, P. (2003). Demographic risk factors, study patterns, and campus



involvement as related to student success among Guilford Technical Community



College students. Community College Review, 31(1), 1-16. Retrieved June 2, 2007



from The WilsonWeb.



Schneider, D., & Dorans, N. (1999). Concordance between SAT I and ACT Scores for



individual students [Electronic Version]. Research Notes, RN-07, 1-8 from



http://www.collegeboard.com/research/pdf/concordance_between_s_10502.pdf.



Sirkin, R. M. (2006). Statistics for the social sciences (3rd ed.). Thousand Oaks,



California: Sage Publications, Inc.



Smith, G. S., & Goetz, E. T. (1988). Using the GED for college placement. College and



University, 63, 360-367.



Soltz, D. F. (1996). The achievements of community college students with GED



certificates: a longitudinal perspective. Community College Journal of Research



and Practice, 20, 269-276.



Spangenberg, G., McClenney, B. & Chisman, F. P. (2005). To ensure America's future:



Building a national opportunity system of adults. Strengthening the links between



Adult Education and community colleges, 2007, 1-56 from



http://www/caalusa.org/ensureamericasfuture.pdf.



Spitzer, T. M. (2000). Predictors of college success: a comparison of traditional and



nontraditional age students. NASPA Journal, 38(1), 82-98. Retrieved June 2, 2007



from The WilsonWeb.

118



Stovall, M. (2000). Using success courses for promoting persistence and completion.



New Directions for Community Colleges (112), 45-54. Retrieved June 2, 2007



from The WilsonWeb.



TAC. (2003). Texas Administrative Code. Chapter 4. Rules Applying to All Public



Institutions of Higher Education in Texas. Subchapter C. Texas Success Initiative.



Retrieved July 15, 2007 from



http://www.thecb.state.tx.us/Rules/tac3.cfm?Chapter_ID=4&Subchapter=C#4.51.



TCALL. (2007). Texas Center for Adult Literacy and Learning. Retrieved July 1, 2007



from http://www-tcall.tamu.edu/provider/search.htm.



TEA. (2007a). Texas Education Agency GED State Examiner's Office. Retrieved July 15,



2007 from http://bass.tea.state.tx.us/TEA.GEDi.Web/Forms/TestCenters.aspx.



TEA. (2007b). Texas Education Agency state assessment policy for Adult Education.



Retrieved July 15, 2007 from http://www-



tcall.tamu.edu/texaslearns/05admanual/1-29-07stateassessmentpolicy.pdf.



THECB. (2000). Closing the gaps. Retrieved April 14, 2007 from



http://www.thecb.state.tx/cfbin/ArchFetch.cfm?DocID=0049&Format=HTML.



Ting, S. R. (1998). Predicting first-year grades and academic progress of college students



of first-generation and low-income families. Journal of College Admission 158,



14-23. Retrieved June 2, 2007 from The WilsonWeb.



Tokpah, C. L., & Padak, N. (2003). Academic challenges. Adult Learning, 14(3), 8-10.



Retrieved April 7, 2007 from The WilsonWeb.



US Census. (2008). State and County Quick Facts. US Census Bureau. Retrieved October



18, 2008, from http://quickfacts.census.gov/qfd/states/36000.html.

119



Wolf, J. C. (1983). Tests of general educational development as a predictor of 2-year



college academic performance. Measurement and Evaluation in Guidance, 16, 4-



12. Retrieved June 2, 2007 from The WilsonWeb.



Wolfe, R. N., & Johnson, S. D. (1995). Personality as a predictor of college performance.



Educational and Psychological Measurement, 55, 177-185. Retrieved June 2,



2007 from The WilsonWeb.



Wyman, F. J. (1997). A predictive model of retention rate at regional two-year colleges.



Community College Review, 25, 29-58. Retrieved June 2, 2007 from The



WilsonWeb.



Zafft, C., Kallenbach, S. & Spohn, J. (2006). Transitioning adults to college: Adult basic



education program models. NCSALL 2007, 1-71 from



http://www.ncsall.net/fileadmin/resources/research/op_collegetransitions.pdf.



Zamani, E. M. (2000). Sources and information regarding effective retention strategies



for students of color. New Directions for Community Colleges (112), 95-104.



Retrieved June 2, 2007 from The WilsonWeb.

APPENDIX A



CONCORDANCE TABLES

121



Table 13



Concordance of GED Reading Test Scores to COMPASS Reading Test Scores



GEDREAD COMPASSREAD

Score Percentile Rank Score Percentile Rank



360 1.1 44 1.1

410 2.2 51 2.2

NA NA 55 3.4

NA NA 56 4.5

NA NA 57 5.6

NA NA 59 6.7

440 8.9 62 7.9

NA NA 63 10.1

NA NA 67 11.2

450 13.3 68 12.4

480 26.7 NA NA

NA NA 73 30.3

490 31.1 74 32.6

500 35.6 76 34.8

510 36.7 77 37.1

NA NA 78 38.2

520 41.1 79 40.4

530 42.2 80 42.7

NA NA 82 46.1

540 48.9 83 48.3

NA NA 84 49.4

550 54.4 85 53.9

560 57.8 86 57.3

NA NA 87 59.6

570 63.3 88 62.9

580 67.8 89 67.4

590 68.9 90 68.5

NA NA 91 71.9

600 73.3 NA NA

610 75.6 92 76.4

620 80.0 93 80.9

630 82.2 NA NA

640 83.3 NA NA

650 84.4 94 85.4

660 86.7 NA NA

670 88.9 NA NA

690 90.0 95 89.9

720 91.1 96 91.0

NA NA 97 93.3

730 96.7 98 97.8

800 100.0 99 100.0

122



Table 14



Concordance of GED Mathematics Test Scores to COMPASS Algebra Test Scores



GEDMATH COMPASSALG

Score Percentile Rank Score Percentile Rank

240 1.1 NA NA

310 2.2 NA NA

330 3.3 NA NA

340 4.4 NA NA

350 6.6 NA NA

370 7.7 NA NA

400 8.8 NA NA

410 14.3 15 3.6

420 17.6 NA NA

430 20.9 NA NA

NA NA 16 26.4

NA NA 18 47.1

460 50.5 NA NA

NA NA 20 54.0

470 57.1 NA NA

NA NA 21 60.9

480 63.7 NA NA

NA NA 22 65.5

490 68.1 23 67.8

440 31.9 17 33.3

NA NA 24 71.3

500 74.7 26 73.6

NA NA 27 78.2

510 80.2 28 80.5

520 81.3 NA NA

540 82.4 NA NA

550 84.6 30 83.9

560 85.7 31 85.1

570 86.8 32 86.2

580 87.9 34 87.4

NA NA 35 88.5

590 90.1 36 90.8

600 92.3 43 92.0

630 93.4 47 93.1

640 94.5 56 94.3

700 95.6 58 95.4

710 96.7 67 96.6

780 97.8 69 97.7

800 100.0 89 100.0


Shared by: yaohongmei
Other docs by yaohongmei
16th Annual
Views: 4  |  Downloads: 0
Pollock_Ethics5e_PPTs_Ch14.ppt - UVU
Views: 3  |  Downloads: 0
incremental backup PPT
Views: 13  |  Downloads: 0
Spine Myelomeningocele
Views: 2  |  Downloads: 0
COMFY CATS
Views: 0  |  Downloads: 0
LDSS Issues
Views: 77  |  Downloads: 1
Strange1-PSE-MEEDM
Views: 0  |  Downloads: 0
Lyssa Junek
Views: 1  |  Downloads: 0
Related docs