140 likes | 257 Views
The Power of Comparison Groups. Rita O’Sullivan Evaluation, Assessment, & Policy Connections School of Education, University of North Carolina at Chapel Hill ( ritao@ unc.edu ) NCCEP Capacity Building Workshop, Las Vegas, NV- February 2013. Important GEAR UP Evaluation Questions:.
E N D
The Power of Comparison Groups Rita O’Sullivan Evaluation, Assessment, & Policy Connections School of Education, University of North Carolina at Chapel Hill (ritao@unc.edu) NCCEP Capacity Building Workshop, Las Vegas, NV- February 2013
Important GEAR UP Evaluation Questions: • Which GEAR UP activities contribute to improved achievement? • Do students who participate in GEAR UP do better academically? • Does GEAR UP participation influence college enrollment and retention?
Possible Evaluation Design Strategies: A. Which GEAR UP activities contribute to improved achievement? • Increase in Percentage/Number of students participating in GEAR UP Activities • Improved achievement from last year • Growth in Pre-test/Post-results
Possible Evaluation Design Strategies: B. Do students who participate in GEAR UP do better academically? • Improved achievement from last year • Growth in Pre-test/Post-results • Comparison of GEAR UP students with non-GEAR UP students in a school
Possible Evaluation Design Strategies: C. Does GEAR UP participation influence college enrollment and retention? • Comparison of GEAR UP students with previous cohort in a school • Comparison of GEAR UP students with school district average • Comparison of GEAR UP schools with non-GEAR UP schools within a district
Possible Alternative Explanations for Evaluation Findings(i.e., Threats to Internal Validity) Evaluation findings vary in how convincing they are due to design strategies usedwith accompanying possible alternative explanations for results:
Key Threats for the Designs: • Which GEAR UP activities contribute to improved achievement? • Do students who participate in GEAR UP do better academically? • Does GEAR UP participation influence college enrollment and retention?
EQUIVALENT Comparison Designs: • EQUIVALENT Groups directly address Sampling threats and indirectly: • History (i.e. services comparison group students received) can still be a problem
Three Credible Comparison Group Design Options: • Cohort Analysis • Multiple Variable Matching • Propensity Score Matching
Cohort Analysis • Definition: Comparing one GEAR UP cohort with a previous cohort within the same school that did not experience GEAR UP • Strengths: • Controls for school differences • Cautions: • One cohort is not necessarily the same as another • One year is not necessarily the same as the next
Multiple Variable Matching • Definition: Creating a “comparable” comparison group based on a variety of relevant variables (e.g., previous achievement, ethnicity, SES, etc.) by setting selection criteria • Strengths: • Matching on multiple variables will provide relatively more equivalent groups for comparison • Challenges: • Need matching data on comparison group • Don’t necessarily know what services selected students have received.
Propensity Score Matching • Definition: Statistically creating a “comparable” comparison group by generating a propensity using a variety of relevant variables (e.g., previous achievement, ethnicity, SES, etc.) • Strengths: • Large number of matching variables make the identification of an “equivalent” group more possible • Cautions: • Need matching data on comparison group • Don’t necessarily know what services selected students have received.
Thank you!Look for a Propensity Score Workshop at NCCEP at July annual meeting in San Francisco