1/18/2019 1:17:10 AM1/18/2019 1:17:10 AM Discussion of “Strategies for Studying Educational Effectiveness” Mark Dynarski Society for Research on Educational.

Slides:



Advertisements
Similar presentations
Copyright © 2010, 2007, 2004 Pearson Education, Inc. Chapter 13 Experiments and Observational Studies.
Advertisements

Jeff Beard Lisa Helma David Parrish Start Presentation.
Math Candel Maastricht University. 1.Internal validity Do your conclusions reflect the “true state of nature” ? 2.External validity or generalizability.
Agenda: Block Watch: Random Assignment, Outcomes, and indicators Issues in Impact and Random Assignment: Youth Transition Demonstration –Who is randomized?
Experiments and Observational Studies.  A study at a high school in California compared academic performance of music students with that of non-music.
Non Experimental Design in Education Ummul Ruthbah.
Early Childhood Education The Research Evidence Deborah Lowe Vandell December 11, 2003.
Student Engagement Survey Results and Analysis June 2011.
Copyright © 2010 Pearson Education, Inc. Chapter 13 Experiments and Observational Studies.
Experiments and Observational Studies. Observational Studies In an observational study, researchers don’t assign choices; they simply observe them. look.
Copyright © 2010 Pearson Education, Inc. Slide
Chapter 13 Observational Studies & Experimental Design.
Copyright © 2007 Pearson Education, Inc. Publishing as Pearson Addison-Wesley Chapter 13 Experiments and Observational Studies.
Striving to Link Teacher and Student Outcomes: Results from an Analysis of Whole-school Interventions Kelly Feighan, Elena Kirtcheva, and Eric Kucharik.
The Evaluation of Charter School Impacts June 30, 2010 Presentation at the 2010 IES Research Conference Philip Gleason ● Melissa Clark Christina Clark.
Slide 1 Estimating Performance Below the National Level Applying Simulation Methods to TIMSS Fourth Annual IES Research Conference Dan Sherman, Ph.D. American.
Slide 13-1 Copyright © 2004 Pearson Education, Inc.
Chapter 5: Producing Data “An approximate answer to the right question is worth a good deal more than the exact answer to an approximate question.’ John.
Selecting and Recruiting Subjects One Independent Variable: Two Group Designs Two Independent Groups Two Matched Groups Multiple Groups.
Propensity Score Matching for Causal Inference: Possibilities, Limitations, and an Example sean f. reardon MAPSS colloquium March 6, 2007.
AFRICA IMPACT EVALUATION INITIATIVE, AFTRL Africa Program for Education Impact Evaluation David Evans Impact Evaluation Cluster, AFTRL Slides by Paul J.
Applying impact evaluation tools A hypothetical fertilizer project.
Non-experimental methods Markus Goldstein The World Bank DECRG & AFTPM.
Africa Impact Evaluation Program on AIDS (AIM-AIDS) Cape Town, South Africa March 8 – 13, Steps in Implementing an Impact Evaluation Nandini Krishnan.
Lincoln Community Learning Centers A system of partnerships that work together to support children, youth, families and neighborhoods. CLC.
Framework of Preferred Evaluation Methodologies for TAACCCT Impact/Outcomes Analysis Random Assignment (Experimental Design) preferred – High proportion.
Using Propensity Score Matching in Observational Services Research Neal Wallace, Ph.D. Portland State University February
Randomized Assignment Difference-in-Differences
Effectiveness of Selected Supplemental Reading Comprehension Interventions: Impacts on a First Cohort of Fifth-Grade Students June 8, 2009 IES Annual Research.
Does Parent Involvement Really Make a Difference? Laura Chianese.
Patricia Gonzalez, OSEP June 14, The purpose of annual performance reporting is to demonstrate that IDEA funds are being used to improve or benefit.
Issues in Selecting Covariates for Propensity Score Adjustment William R Shadish University of California, Merced.
Copyright © 2009 Pearson Education, Inc. Chapter 13 Experiments and Observational Studies.
Experimental Design. Some Definitions Observational Study –Observes outcomes as they occur without imposing any treatment Experiment –Actively imposes.
SESRI Workshop on Survey-based Experiments
CHAPTER 4 Designing Studies
Constructing Propensity score weighted and matched Samples Stacey L
Lindsay K. Lightner and Judith A. Morrison Washington State University
Introduction The two-sample z procedures of Chapter 10 allow us to compare the proportions of successes in two populations or for two treatments. What.
Sec 9C – Logistic Regression and Propensity scores
Experimental Design-Chapter 8
Quasi-Experimental Methods
SESRI Workshop on Survey-based Experiments
John Loucks St. Edward’s University . SLIDES . BY.
Experiments: What Can Go Wrong?
Experiments and Observational Studies
March 2017 Susan Edwards, RTI International
School Climate Data Workshop
S1316 analysis details Garnet Anderson Katie Arnold
Impact evaluation: The quantitative methods with applications
Matching Methods & Propensity Scores
Matching Methods & Propensity Scores
EMPIRICAL STUDY AND FORECASTING (II)
Impact Evaluation Methods
Matching Methods & Propensity Scores
Chapter 4: Designing Studies
Statistical Reasoning December 8, 2015 Chapter 6.2
Impact Evaluation Methods: Difference in difference & Matching
Evaluating Impacts: An Overview of Quantitative Methods
Sampling and Power Slides by Jishnu Das.
Health & Physical Education Teacher Education School of Kinesiology
Madison Elementary / Middle School and the New Accountability System
Heidi H. Erickson Jay P. Greene, Angela R. Watson, and Molly I. Beck
Milwaukee Public Schools University of Wisconsin-Milwaukee
Chapter 4: Designing Studies
Warmup A drug company wants to test a new drug, statsium, to determine if it improves memory in young adults. There will be two treatments: a daily.
Sample Sizes for IE Power Calculations.
Steps in Implementing an Impact Evaluation
10/28/ B Experimental Design.
Heidi H. Erickson Jay P. Greene, Angela R. Watson, and Molly I. Beck
Presentation transcript:

1/18/2019 1:17:10 AM1/18/2019 1:17:10 AM Discussion of “Strategies for Studying Educational Effectiveness” Mark Dynarski Society for Research on Educational Effectiveness Annual Meetings Landsdowne, VA December 2006 Roadmap/overview of what we’re going to talk about today. Begin with impact estimates for elementary schools In 1st report presented impact estimates for original set of 7 elementary grantees. Since then added 5 more elem. grantees. Today we’ll present impact estimates for full set of 12 elementary grantees in year 1, which is the first year after random assignment. Next we’ll present year 2 middle school outcome differences.

Themes 1/18/2019 1:17:10 AM1/18/2019 1:17:10 AM Alternatives to study effectiveness in situations where experiments may not be feasible or desirable Issues Focus on experiments should not be lost Prospective versus retrospective studies Costs and practicality A few notes before turning to elementary findings: Random assignment successful at aligning T and C groups. Next slide presents sample info. All estimates regression adjusted to improve precision of estimates. Focus on findings significant at 5% or better. Brief overview of findings here followed by more detail in following slides. Overall, impact findings changed little when we added the 5 new sites. Found moderate program attendance -- students attended ~ 2 days/week on average. Program affected students’ location and supervision after school. The program: Reduced care by parents and siblings Increased care by other adults Had no effect on self care. Multiple definitions used – all yielded same results. Reduced % of students in own homes after school Increased % of students at school after school On academic outcomes, found that: No improvement in grades or reading test scores Teachers were less likely to report that students often completed their homework Impact and T/C means similar to those from yr 1 but this year the result is significant. Additional sample reduced variability of estimate. Students reported feeling safer after school. Few impacts on developmental outcomes. Program increased % of students reporting that they helped others after school. No effect on other developmental outcomes such as working on teams, sticking to beliefs when others disagree, getting along with others. No effect on behavior outcomes such as suspensions, being sent to office, being disciplined in class. Some measures of parental involvement increased for program participants relative to control students. Also found parents of participants were more likely to be in the labor force.

Regression Discontinuities 1/18/2019 1:17:10 AM1/18/2019 1:17:10 AM + Being able to allocate services according to need is attractive - Less power and stronger validity conditions than experiments, does not support subgroup estimation as well as experiments Retrospective: if experiment is not feasible, is a score available? Can it be gamed? Prospective: If thinking about an RD study, the experiment may be a better choice (use risk strata) A few notes before turning to elementary findings: Random assignment successful at aligning T and C groups. Next slide presents sample info. All estimates regression adjusted to improve precision of estimates. Focus on findings significant at 5% or better. Brief overview of findings here followed by more detail in following slides. Overall, impact findings changed little when we added the 5 new sites. Found moderate program attendance -- students attended ~ 2 days/week on average. Program affected students’ location and supervision after school. The program: Reduced care by parents and siblings Increased care by other adults Had no effect on self care. Multiple definitions used – all yielded same results. Reduced % of students in own homes after school Increased % of students at school after school On academic outcomes, found that: No improvement in grades or reading test scores Teachers were less likely to report that students often completed their homework Impact and T/C means similar to those from yr 1 but this year the result is significant. Additional sample reduced variability of estimate. Students reported feeling safer after school. Few impacts on developmental outcomes. Program increased % of students reporting that they helped others after school. No effect on other developmental outcomes such as working on teams, sticking to beliefs when others disagree, getting along with others. No effect on behavior outcomes such as suspensions, being sent to office, being disciplined in class. Some measures of parental involvement increased for program participants relative to control students. Also found parents of participants were more likely to be in the labor force.

Propensity Scoring: Observations 1/18/2019 1:17:10 AM1/18/2019 1:17:10 AM It’s an interesting issue to debate It’s useful to have hundreds of undergrads to randomly assign If you are on the other side of a statistical argument with Don Rubin, you should reconsider A few notes before turning to elementary findings: Random assignment successful at aligning T and C groups. Next slide presents sample info. All estimates regression adjusted to improve precision of estimates. Focus on findings significant at 5% or better. Brief overview of findings here followed by more detail in following slides. Overall, impact findings changed little when we added the 5 new sites. Found moderate program attendance -- students attended ~ 2 days/week on average. Program affected students’ location and supervision after school. The program: Reduced care by parents and siblings Increased care by other adults Had no effect on self care. Multiple definitions used – all yielded same results. Reduced % of students in own homes after school Increased % of students at school after school On academic outcomes, found that: No improvement in grades or reading test scores Teachers were less likely to report that students often completed their homework Impact and T/C means similar to those from yr 1 but this year the result is significant. Additional sample reduced variability of estimate. Students reported feeling safer after school. Few impacts on developmental outcomes. Program increased % of students reporting that they helped others after school. No effect on other developmental outcomes such as working on teams, sticking to beliefs when others disagree, getting along with others. No effect on behavior outcomes such as suspensions, being sent to office, being disciplined in class. Some measures of parental involvement increased for program participants relative to control students. Also found parents of participants were more likely to be in the labor force.

Propensity scoring: Shadish and Clark 1/18/2019 1:17:10 AM1/18/2019 1:17:10 AM PS is based on a theorem and it has to work if assumptions are met Small bias (10%); unclear how scoring would perform with large bias Prospective: can be more expensive than an experiment (Rubin recomends potential comparison group 3 to 4 times larger than matched group) Retrospective: probably do not have data to model selection A few notes before turning to elementary findings: Random assignment successful at aligning T and C groups. Next slide presents sample info. All estimates regression adjusted to improve precision of estimates. Focus on findings significant at 5% or better. Brief overview of findings here followed by more detail in following slides. Overall, impact findings changed little when we added the 5 new sites. Found moderate program attendance -- students attended ~ 2 days/week on average. Program affected students’ location and supervision after school. The program: Reduced care by parents and siblings Increased care by other adults Had no effect on self care. Multiple definitions used – all yielded same results. Reduced % of students in own homes after school Increased % of students at school after school On academic outcomes, found that: No improvement in grades or reading test scores Teachers were less likely to report that students often completed their homework Impact and T/C means similar to those from yr 1 but this year the result is significant. Additional sample reduced variability of estimate. Students reported feeling safer after school. Few impacts on developmental outcomes. Program increased % of students reporting that they helped others after school. No effect on other developmental outcomes such as working on teams, sticking to beliefs when others disagree, getting along with others. No effect on behavior outcomes such as suspensions, being sent to office, being disciplined in class. Some measures of parental involvement increased for program participants relative to control students. Also found parents of participants were more likely to be in the labor force.

Using the methods to enhance experiments 1/18/2019 1:17:10 AM1/18/2019 1:17:10 AM Propensity scoring or HLM for dosage HLM for variable counterfactuals RD for testing model variations within a treatment group A few notes before turning to elementary findings: Random assignment successful at aligning T and C groups. Next slide presents sample info. All estimates regression adjusted to improve precision of estimates. Focus on findings significant at 5% or better. Brief overview of findings here followed by more detail in following slides. Overall, impact findings changed little when we added the 5 new sites. Found moderate program attendance -- students attended ~ 2 days/week on average. Program affected students’ location and supervision after school. The program: Reduced care by parents and siblings Increased care by other adults Had no effect on self care. Multiple definitions used – all yielded same results. Reduced % of students in own homes after school Increased % of students at school after school On academic outcomes, found that: No improvement in grades or reading test scores Teachers were less likely to report that students often completed their homework Impact and T/C means similar to those from yr 1 but this year the result is significant. Additional sample reduced variability of estimate. Students reported feeling safer after school. Few impacts on developmental outcomes. Program increased % of students reporting that they helped others after school. No effect on other developmental outcomes such as working on teams, sticking to beliefs when others disagree, getting along with others. No effect on behavior outcomes such as suspensions, being sent to office, being disciplined in class. Some measures of parental involvement increased for program participants relative to control students. Also found parents of participants were more likely to be in the labor force.