Professor Kathy

Document Sample
Professor Kathy Powered By Docstoc
					RESEARCH
EXAMPLES
Professor Kathy
          Professor Kathy’s Design
             Aim and Question
Does classical background music (CBM), easy listening
  background music (ELBM), or no background music
  (NBM) produce faster and/or more correct coding?


IDV = presence and type of music
DV = speed of coding (Question 1) and
     accuracy of coding (Question 2)
                Professor Kathy
Question and Hypotheses:

Question 1:
Q1: Do CBM, ELBM, and NBM coders have different
     speeds of coding?
H1: CBM, ELBM, and NBM coders will have different rates
     of files coded per hour.

Question 2:
Q1: Are CBM, ELBM, and NBM coders equally accurate?
H1: CBM, ELBM, and NB coders will have different
     accuracy rates as determined by file reviewers.
                  Professor Kathy
Design: True experimental design

Population of Interest, Unit of Analysis, and Sampling:
  Population is people who code for university researchers.
  Assigning subjects: Coder Training in September using
  standard procedures. Coders must pass a test. Coders are
  then randomly assigned to one of three conditions for study.
  Unit of Analysis: Entries are made every 30 minutes for one
  day using coding sheets.
                 Professor Kathy
Constructs and Variables:
Music           Number of coded files/hour
                Number of mistakes made/hour

Control:
• Random assignment of coders
• Accuracy of coding checked by “blind” raters
• Individual cubicles for coders
• Random daily rotation of music in each cubicle to address
     environmental differences
               Professor Kathy
Strengths and Limitations of Design:
Internal validity – excellent
External validity – May be limited. May not be able
   to generalize findings to other coders doing other
   kinds of coding.

Time Line:
Study is expected to go quickly
                Professor Kathy
IRB Issues:
Options are:
• Research without consent?
• Tell people they are being studied in some way without
      telling them about the music? (MILD DECPTION)
• Inform subjects what is going on (NO DECEPTION)

IRB elected second option in this case.
Review sample consent form on Page 174, Figure 9.2.
                  Professor Kathy
•   No pilot study – had used coder in past
•   Set-up for music while coding
•   Created codebook.
•   Set up Excel spreadsheet with coders names.
•   Wrote a SAS program “KATHY1.SAS” .
•   Imported spreadsheet into SAS and randomly assigned
    coders to groups using SAS codes.
•   Spreadsheet then exported to Excel for data entry. ID
       numbers and group numbers for coders brought
       forward to spreadsheet instead of names.
                   Professor Kathy
•   Coders finished files were given to research assistants to
       check
•   Random assignment to quality assurance coders for
       checking
•   Files were checked by three reviewers for accuracy. Last
       reviewer entered the number of files completed and the
       number of errors on paper. Professor Kathy entered
       information into Excel spreadsheet. Name of
       spreadsheet changed to KATHYDATA02.XLS
•   Spreadsheet was imported back into SAS for analysis
                 Professor Kathy
•   Total files over five sessions (FALL), total number
      of errors (EFALL), average number of errors
      over all sessions (EPFALL), basic numbers
      (PROC, FREQ, PRINT, and UNIVARIATE).
•   KATH2.SAS written to do data management tasks
MARIA
                  Maria’s Design
                 Aim and Question
Are higher residential density and poverty associated with
   increases in homicide and suicide rates in zip codes in San
   Diego?
Question 1A: Is higher residential density associated with
              higher homicide rates?
Hypothesis 1A: Higher residential density (2000 Census) will
  be correlated with a higher homicide rate (2000-2001) in
  San Diego zip codes.
Question 1B: Is lower income associated with higher
              homicide rates?
Hypothesis 1B: Lower income (2000 Census) will be
  correlated with a higher homicide rates (2000-2001) in San
  Diego zip codes.
                Maria’s Design
               Aim and Question
Question 2A: Is higher residential density associated
  with higher suicide rates?
Hypothesis 2A: Higher residential density (2000
  Census) will be correlated with higher suicide rates
  (2000-2001) in San Diego zip codes.

Question 2B: Is lower income associated with higher
  suicide rates?
Hypothesis 2B: Lower income will be correlated with
  higher suicide rates (2000-2001) in San Diego zip
  codes.
               Maria’s Design
              Aim and Question
Question 3: What other community factors are
  associated with homicide rates?

Question 4: What other community factors are
  associated with suicide rates?
                  Maria’s Design
Type of design:
• One group design
• Nonexperimental design
• Questions 1 & 2 are testing a theoretically derived
     relationship; Questions 3 & 4 are more descriptive or
     exploratory, and do not require hypotheses.

Population of Interest, Unit of Analysis & Sampling
Sampling frame and unit of analysis is the zip code.
Data will cover 2 years
Exclude zip codes with <5000 people????
                  Maria’s Design
Constructs and Variables:
Hypotheses 1 and 2 variables:
  # of homicides per 10,000 residents
  # of suicides pr 10,000 residents
  Residential density in zip code
  Poverty in zip code

Maria made changes to measure:
• Percentage of all residences that are single person occupied
• Number of people in zip code divided by square miles in
      zip code
                  Maria’s Design
                Revised Hypotheses
H1A1: More residents per square mile (2000 Census) will be
  correlated with a higher homicide rate (2000-2001) in San
  Diego zip codes.
H1A2: Higher percentages of single-person residences (2000
  Census) will be correlated with a higher homicide rate
  (2000-2001) in San Diego zip codes.
H2A1: More residents per square mile (2000 Census) will be
  correlated with a higher suicide rate (2000-2001) in San
  Diego zip codes.
H2A2: Higher percentages of single-person residences (2000
  Census) will be correlated with a higher suicide rate (2000-
  2001) in San Diego zip codes.
               Maria’s Design
Maria also added Age, Race (Whites, Blacks,
  Hispanics, Asians), and other factors that might
  influence suicide rates. Income will be determined
  by median income in the zip code.
                     Maria’s Design
Control: More hypotheses are added to provide for spurious causality.
H5A1 (multivariate version of H1A1):
  More residents per square mile (2000 Census) will be correlated with
  a higher homicide rate (2000-2001) in San Diego zip codes while
  controlling for other factors that are associated with homicide rates.
H5A2 (multivariate version of H1A2):
  Higher percentages of single-person residences (2000 Census) will
  be correlated with a higher homicide rate (2000-2001) in San Diego
  zip codes while controlling for other factors that are associated with
  homicide rates.
H5B (multivariate version of H1B): Lower income will be correlated
  with higher suicide rates (2000-2001) in San Diego zip codes while
  controlling for other factors that are associated with homicide rates.
                     Maria’s Design
H6A1 (multivariate version of H2A1):
  More residents per square mile (2000 Census) will be correlated with
  a higher suicide rate (2000-2001) in San Diego zip codes while
  controlling for other factors that are associated with homicide rates.

H6A2 (multivariate version of H2A2):
  Higher percentages of single-person residences (2000 Census) will
  be correlated with a higher suicide rate (2000-2001) in San Diego zip
  codes while controlling for other factors that are associated with
  homicide rates.
H6B1 (multivariate version of H2B1):
  Lower income (2000 Census) will be correlated with a higher suicide
  rate (2000-2001) in San Diego zip codes while controlling for other
  factors that are associated with homicide rates.
.
                Maria’s Design
Strengths and Limitations of Design:
+Based on prior work; Good theoretical basis
+Finding will result from a large number of zip codes
-Able to test only association NOT causality
-Ability to sample a rare event may throw off findings
Suicides/Homicides may not have occurred in the zip
   code where person lived (Research shows most
   suicides and homicides occur near home, so this
   will be included in the write-up.)
                Maria’s Design
Time Line:
“cushion time” built in

IRB Issues:
“Exempt” status. No identifying data. Using publicly
  available information.
                        Maria
•   Difficulty in obtaining census data –finally got
      DVD from school for census data
•   Difficulty in obtaining measure of land in each zip
      code to determine population density-finally
      obtained data from government resource
•   Downloaded data into Excel file, printed out
      acreage and coroner data and entered these by
      hand into the Excel file
                         Maria
Problems with units of analysis:
•  Individual vs Household level data
•  Census data is at the family or household level
•  To measure at the individual level
•  Wanted to measure median individual income, so had 3
   options:
      1) Use census variable to measure individual mean
            income (problem: skewing due to outliers)
      2) Use census variable to measure household median
            income (problem: bias if some zip codes had
            larger average household sizes)
      3) Use percentage of persons in each zip code who
            were in poverty
                         Maria
Codebook construction:
• Tracked variable name and source variable code
• Listed each variable twice, once listing the source, range
     and missing data codes for each variable and once
     listing a more heuristic definition
• Designed so that every variable can be traced back to the
     original source data easily
• An SAS program was written to input the data from the
     codebook and transform them into the variables needed.
• Basic data management completed. Ready for analyis.
ABIGAIL
            Abigail’s Design
      Aim, Question, and Hypotheses
What is the degree to which work climate factors
 (supportive and conflictual communication) and
 nonwork supports (family and friend emotional
 support, social support, and advice/guidance) are
 associated with the impact of client death on
 workers (postevent intrusive and avoidance
 experiences).
                 Abigail’s Design
Q1: Are organizational factors (supportive and conflictual
  communication) associated with the impact (intrusiveness,
  avoidance) of client fatalities?

H1A1: Organizational supportive communication will be
  associated with lower levels of worker-reported postfatality
  event intrusive experiences.

H1A2: Organizational conflictual communication will be
  associated with higher levels of worker-reported
  postfatality event intrusive reactions.
                 Abigail’s Design
Q2: Are nonwork social support factors (emotional, social,
  advice/guidance) associated with the impact (intrusiveness,
  avoidance) of client fatalities?
H2A1: Emotional support by friends will be associated with
  lower levels of worker-reported postfatality event intrusive
  experiences.
H2A2: Emotional support by friends will be associated with
  lower levels of worker-reported postfatality event
  avoidance reactions.
H2A3: Emotional support by family will be associated with
  lower levels of worker-reported postfatality event intrusive
  experiences.
H2A4: Emotional support by family will be associated with
  lower levels of worker-reported postfatality event
  avoidance reactions.
                 Abigail’s Design
H2B1: Social supports by friends will be associated with
  lower levels of worker-reported postfatality event intrusive
  experiences.
H2B2: Social supports by friends will be associated with
  lower levels of worker-reported postfatality event
  avoidance reactions.
H2B3: Social supports by family will be associated with lower
  levels of worker-reported postfatality event intrusive
  experiences.
H2B4: Social supports by family will be associated with lower
  levels of worker-reported postfatality event avoidance
  reactions.
                 Abigail’s Design
H2C1: Advice and guidance by friends will be associated
  with lower levels of worker-reported postfatality event
  intrusive experiences.
H2C2: Advice and guidance by friends will be associated
  with lower levels of worker-reported postfatality event
  avoidance reactions.
H2C3: Advice and guidance by family will be associated with
  lower levels of worker-reported postfatality event intrusive
  experiences.
H2C4: Advice and guidance by family will be associated
  with lower levels of worker-reported postfatality event
  avoidance reactions.
               Abigail’s Design
Q3: What other personal characteristics are associated
  with increase in the impact of child fatalities?

Q4: Do effects of Q1 and Q2 persist in the presence
  of controls for personal characteristics?

Type of Design:
• One group
• Nonexperimental and correlational
• Mainly testing theoretically derived hypotheses
               Abigail’s Design
Population of Interest, Unit of Analysis, Sampling:
Will begin by using fatality review commission
  reports to identify how many children died in the
  last year and were open for services. Then plans to
  locate workers and administer survey packets to
  them. Will be testing entire population of interest
  in the state.
               Abigail’s Design
Constructs and Variables:
Independent constructs:
   The impact of the fatality, organizational
     communication, support by friends and family
     and other workers (i.e., age, experience, etc.)

Dependent variables:
  Worker factors (descriptive)
  Variables drawn from subscales (intrusiveness and
    avoidance)
               Abigail’s Design
Control:
Control for worker characteristics in the multivariate
  model in Q4 only.

Strengths and Limitations of Design:
+ She has an entire State to draw sample from
+ Measures appear to be good matches to her
     constructs
+ She is studying an important and unexplored area
               Abigail’s Design
- Sample size and response rate are unable to be
      determined presently.
- Differing time frames between the event and the
      survey for each worker
- Content validity – Is she studying all the key
      issues?
Time Line: Plans to do the study quickly
IRB Issues: Surveys by mail so unable to monitor
  reactions. Phone counseling? Internal counseling
  resources listed on the consent form?
                          Abigail
•   Review of annual report and child fatalities
•   Met with representative of research division of agency
•   Human subjects packet
    • Cover page
    • Consent form
    • General Information form
    • SSB, OCS, IES Scales
•   Developed mailout packet
    • Social Support Behaviors Scale (SSB)
    • Impact of Events Scale (IES)
    • Organizational Climate Scale (OCS)
    • Letter introducing project
    • Demographic questionnaire
    • Self-addressed return envelope
                           Abigail
•   Met with crisis hotline and state child welfare agency reps
       and obtained list of workers that had fatalities on their
       caseloads in last 12 months
•   Received IRB approval
•   Sent out packets to workers; follow-up phone call one
       week after one week after mailout to answer questions
       workers may have
•   Data was entered into Excel spreadsheet and commands
       labeled, then imported into SAS
•   Command written to create subscale totals
•   Basic data management completed.
YUAN
              Yuan’s Design
       Aim, Question, and Hypotheses
Wants to determine if cognitive behavioral therapy (CBT) is
  more effective than standard therapy (ST) in reducing male
  domestic violence
Working with an agency that runs groups for males who
  batter. Agency has been using “regular” treatment, but
  wants to try CBT.
Agency works with court-mandated clients and has close
  relationship with law enforcement. Police report any
  domestic violence reports made against clients before,
  during, and for two years after treatment.
                Yuan’s Design
Q1: Do subjects receiving CBT and ST show different
  changes in AS subscale scores?
H1A1: Subjects receiving CBT and ST will show
  different levels of change in the AS avoidance
  subscale from pretest to posttest (change during
  treatments).
H1A2: Subjects receiving CBT and ST will show
  different levels of change in the AS avoidance
  subscale from posttest to six-months posttreatment
  (change following treatment).
                Yuan’s Design
H1B1: Subjects receiving CBT and ST will show
  different levels of change in the AS empathy
  subscale from pretest to posttest (change during
  treatments).
H1B2: Subjects receiving CBT and ST will show
  different levels of change in the AS empathy
  subscale from posttest to six-months posttreatment
  (change following treatment).
                Yuan’s Design
H1C1: Subjects receiving CBT and ST will show
  different levels of change in the AS forethought
  subscale from pretest to posttest (change during
  treatments).
H1C2: Subjects receiving CBT and ST will show
  different levels of change in the AS forethought
  subscale from posttest to six-months posttreatment
  (change following treatment).
                Yuan’s Design
H1D1: Subjects receiving CBT and ST will show
  different levels of change in the AS usefulness
  subscale from pretest to posttest (change during
  treatments).
H1D2: Subjects receiving CBT and ST will show
  different levels of change in the AS usefulness
  subscale from posttest to six-months posttreatment
  (change following treatment).
                Yuan’s Design
Q2: Do subjects receiving CBT and ST show different
  rates of reported violence?
H2: Subjects receiving CBT and ST will show
  different rates of recidivism during the six months
  following treatment.
Q3: Do empathy and argument avoidance predict
  recidivism?
H3A: People with higher AS avoidance scores at
  posttest will have lower rates of recidivism.
                Yuan’s Design
Q3: Do empathy and argument avoidance predict
  recidivism?
H3A: People with higher AS avoidance scores at
  posttest will have lower rates of recidivism.
H3B: People with higher AS empathy scores at
  posttest will have lower rates of recidivism.
H3C: People with higher AS forethought scores at
  posttest will have lower rates of recidivism.
H3D: People with higher AS usefulness scores at
  posttest will have lower rates of recidivism.
                 Yuan’s Design
Type of Design:
Quasi-experimental design because he is not able to do
  random assignment.
H1A1, H1B1, H1C1, and H1D1 are pretest/posttest
  design
H1A2, H1B2, H1C2,and H1D2 relate to a standard
  follow-up design
H3 is a correlational design. (Each group should be
  tested separately to see if the effect occurs for both
  experimental and comparison samples.)
                    Yuan’s Design
Constructs and Variables:
Independent variables:
• CBT versus ST treatment group membership (Q1and Q2)
• Subscale scores (Q3)

Dependent Variables:
AS empathy, avoidance, forethought, and usefulness scores
  posttest and follow-up (Q1)
Recidivism (Q2 and Q3)

Control Variables:
Age, marital status, prior reports to police
                   Yuan’s Design
Control:
Use of comparison group and inclusion of control variables in
   multivariate models
Similarity of control and experimental groups with regard to
   demographics ad initial subscale scores

Strengths and Limitations of Design:
- Lacks random assignment
- Difficulty defining what “standard treatment” is
- Questions about internal validity
- Questions about treatment fidelity
- CBT therapists have more training
                   Yuan’s Design
Time Line:
Very compressed schedule to collect, analyze and write data
  and results

IRB Issues:
Using well recognized treatment protocol with subjects who
   are mandated to receive treatment, but subjects may opt out
   of the study.
Consent form does not state which group the subject will be
   in.
Statement was added to give consent for follow-up data to be
   collected from agency and law enforcement or the next
   year.
                        Yuan
Needed to distribute one consent form
Administered the argument scale (fictional scales) at
   the first and last session himself
Pretest data was obtained, but therapists then refused
   to participate in the posttest and follow-up portions
   of the project. Questions and hypotheses had to be
   revised due to reduced amount of data.
        Yuan’s Revised Questions
Q1: Do subjects receiving CBT show changes in AS subscale
  scores from pretest to posttest?
  (Addresses change during treatment)
  H1A: Subjects will show change in AS avoidance scale
  (pretest-posttest).
  H1B: Subjects will show change in the AS empathy
  subscale (pretest-posttest).
  H1C: Subjects will show change in the AS forethought
  subscale (pretest-posttest).
  H1D: Subjects will show change in the AS usefulness
  subscale (pretest-posttest).
        Yuan’s Revised Questions
Q2: Do subjects receiving CBT show changes in AS subscale
  scores from post test-follow-up?
  (Do changes persist?)
  H2A: Subjects will show change in the AS avoidance
  subscale (posttest-follow-up).
  H2B: Subjects will show change in the AS empathy
  subscale (posttest-follow-up).
  H2C: Subjects will show change in the AS forethought
  subscale (posttest-follow-up).
  H2D: Subjects will show change in the AS usefulness
  subscale (posttest-follow-up).
        Yuan’s Revised Questions
Q3: Do subjects receiving CBT and ST show different
  rates of reported violence?
  H3: Subjects receiving CBT and ST will show
  different rates of recidivism during the six months
  following treatment.

Questions 1 and 2 are correlational.
Question 3 is quasi-experimental and uses a
  comparison group.
                     Yuan
Yuan reversed scored his data set.
Data management program does labeling and subscale
  summing.
Codebook is quite long
Data is ready for analysis.
JOHN
             John’s Design
      Aim, Question, and Hypotheses
What strengths do Bosnian refugees report to be
 helpful to the acculturation process?

Subjects will define their strengths
Use Berry’s model as an orienting framework for
   understanding the acculturation process
No specific questions or hypotheses will be tested
Individual interviews will be used to collect
   information
                   John’s Design
Type of Design:
• Exploratory design – use narrative data from individual
     interviews to generate new ideas
• Use of grounded theory

Population of Interest, Unit of Analysis, Sampling:
• Theoretical sampling to recruit Bosnian refuges with a
     wide range of adaptation outcomes
• Sampling frame: Bosnians using services or attending
     events at the Resettlement House
• Subjects must have been in the US for at least a year
                  John’s Design
• Use of an interpreter present at interviews
• Questionnaire includes open and closed-ended
    questions
• 20 interviews to start and then reassess to see if this
    is enough

Constructs and Variables:
Ecological framework
Inductive study – expectation for constructs and
   variables to emerge from interviews
                   John’s Design
Control: Control is not a focus of the design
Strengths and Limitations of Design:
• Question about the sample and if it is adequately
      broad
• May miss certain groups of assimilated people who      no
   longer use immigrant-specific services
Time Line: Two months for interviews and data analysis
IRB Issues:
• Counselor on call at agency
• Consent form states transcribed tapes will be destroyed
• No names or identifying information will be transcribed
                            John
Recruitment was a problem. Not enough appropriate subjects
  (age and living in the US for a year) responded to posting of
  flyers. (8 subjects recruited)
• Asked all appropriate respondents for name and phone
      numbers of friends who could be interviewed (snowball
      sampling). 12 subjects were recruited in total. (In reality
      this is not sufficient to reach saturation, especially when
      looking into subgroups. Should have 3-5 times as many
      subjects. This is number is used for demonstration
      purposes only).
• Revision of human subjects proposal was necessary due to
      this.
                         John
•   Usefulness of the questionnaire was examined and
      no changes were needed.
•   Each participant’s recorded session was transcribed
      by John.
•   Data analysis is ready to begin.

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:4
posted:1/6/2012
language:
pages:63