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