1. Reviewing systematic reviews: meta- analysis of What Works Clearinghouse computer-assisted reading interventions. October 2012 Improving Education through Accountability and Evaluation: Lessons from Around the World Rome, Italy Andrei Streke ● Tsze Chan

2. Session 4.2: Building and Interpreting Scientific Evidence Thursday, October 4th 2

3.  What Works Clearinghouse (WWC) systematic reviews  Meta-analysis of computer-assisted programs across WWC topic areas, reading outcomes  Meta-analysis of computer-assisted programs within Beginning Reading topic area (grades K-3) Presentation Overview 3

4. What Works Clearinghouse (WWC) is a “central and trusted source of scientific evidence for what works in education.” WWC produces systematic reviews on the effectiveness of educational interventions (programs, curricula, products, and practices) grouped by topic areas. Meta-analysis is a statistical technique that summarizes quantitative findings across similar studies. Each study’s findings are converted to a standard effect size. Computer-assisted interventions encompass reading software products, and programs that combine a mix of computer activities and traditional curriculum elements. Key terms 4

5.  A clearly stated set of objectives with pre-defined eligibility criteria for studies  An explicit reproducible methodology  A systematic search that attempts to identify all studies that would meet the eligibility criteria  An assessment of the validity of the findings of the included studies  A systematic presentation, and synthesis, of the characteristics and findings of the studies WWC Systematic Review 5

6.  Extraction of statistical and descriptive information from intervention reports and study review guides  Aggregation of effect sizes across studies  Moderator Analysis -- ANOVA type -- Regression type Meta-Analysis of Reading interventions 6

7. WWC products:  Intervention reports http://ies.ed.gov/ncee/wwc/publications_reviews.aspx http://ies.ed.gov/ncee/wwc/publications_reviews.aspx  Practice guides  Quick reviews Normative documents ( http://ies.ed.gov/ncee/wwc ) : ( http://ies.ed.gov/ncee/wwc  WWC Procedures and Standards Handbook  WWC topic area review protocol WWC Systematic Review 7

8. 8

9. Meta-analysis of computer-assisted programs across WWC topic areas, reading outcomes  Does the evidence in WWC reports indicate that computer-assisted programs increase student reading achievement? 9

10. Computer-assisted interventions 10

11.  Earobics ® is interactive software that provides students in pre-K through third grade with individual, systematic instruction in early literacy skills as students interact with animated characters. The program builds children’s skills in phonemic awareness, auditory processing, and phonics, as well as the cognitive and language skills required for comprehension. Example of computer-assisted programs 11

12.  Effect Sizes  Aggregation Method  Testing for Homogeneity  Fixed and Random Effects Models  Moderator Analysis -- ANOVA type -- Regression type Meta-Analysis procedures 12

13. Effect Size 13 (1) Effect size (Hedges & Olkin, 1985):

14. Flowchart for calculation of effect size (Tobler et al., 2000) 14

15. Number of students and effect sizes by topic area 15 Type of Program Number of students Number of effect sizes totalinterventioncontrol Adolescent Literacy 26970127171425359 Beginning Reading 263613391297151 Early Childhood Education 91044746339 English Language Learners 3081731356 Total308241467616148255

16. Aggregation of Effect Sizes 16 (1) Effect size (Hedges): (2) Effect size variance: (3) Weighted average effect size: (4) Weighted average effect size variance: Weight (w)= (Variance) -1

17.  Fixed effects model weights each study by the inverse of the sampling variance.  Random effects model weights each study by the inverse of the sampling variance plus a constant that represents the variability across the population effects (Lipsey & Wilson, 2001). Fixed and Random Effects Model weights 17 This is the random effects variance component.

18. Computer-assisted programs, random effects 18 WWC Topic Area Number of Studies Weighted Effect Size Standard Error Lower Confidence Interval Upper Confidence Interval Z- value P- value Adolescent literacy 310.130.030.070.184.560.00 Beginning reading 330.280.060.160.404.710.00 Early childhood education 60.120.07-0.010.251.740.14 English language learners 30.300.27-0.230.831.110.38

19. Computer-assisted reading interventions, topic area effects and 95% CIs 19 0.13 0.28 0.12 0.30 -0.4 -0.2 0 0.4 0.6 0.8 1 Adolescent Literacy Beginning ReadingEarly Childhood Education English Language Learners

20. Meta-analysis of computer-assisted programs within Beginning Reading topic area  Are computer-assisted reading programs more effective than non-computer reading programs in improving student reading achievement? 20

21.  Manuscript is written in English and published 1983 or later  Both published and unpublished reports are included  Eligible designs: RCT; QED with statistical controls for pretest and/or a comparison group matched on pretest; regression discontinuity; SCD  At least one relevant quantitative outcome measure  Manuscript focuses on beginning reading  Focus is on students ages 5-8 and/or in grades K-3  Primary language of instruction is English Selection Criteria for Beginning Reading Topic Area 21

22. Beginning Reading Topic Area 22

23.  Reading Recovery® is a short-term tutoring intervention intended to serve the lowest- achieving first-grade students. The goals of Reading Recovery® are to promote literacy skills, reduce the number of first-grade students who are struggling to read, and prevent long-term reading difficulties. Reading Recovery® supplements classroom teaching with one-to-one tutoring sessions, generally conducted as pull-out sessions during the school day. Example of “other” reading programs 23

24. Number of students and effect sizes by type of program: Beginning Reading topic area 24 Type of Program Number of studentsNumber totalinterventioncontrol of effect sizes BR Computer-Assisted Programs263613391297151 Other BR Programs759140423549174 Total Beginning Reading1022753814846325

25. Mathematica ® is a registered trademark of Mathematica Policy Research. Beginning Reading programs, random effects 25 Type of ProgramnM Standard Error 95% Lower 95% Upper Z-value P- value Computer-assisted programs 330.280.060.160.404.710.000 Other BR programs470.390.040.320.479.840.000 Beginning Reading Total 800.350.030.290.4210.650.000

26. Beginning Reading Interventions, Random Effects, 95% Confidence Intervals 26

27. Modeling between study variability:  Categorical models (analogous to a one-way ANOVA)  Regression models (continuous variables and/or multiple variables with weighted multiple regression) Moderator Analysis, random effects 27

28.  Population  Design  Sample size  Control group  Reading domain Categorical analysis: moderators of program effectiveness 28

29. Weighted mean Effect Sizes for moderators: 80 studies, Beginning Reading, random effects 29

30. Weighted mean Effect Sizes for moderators: 80 studies, Beginning Reading, random effects 30

31. Dummy Variables for Regressions 31

32. Regression Statistics for BR Programs, Random effects 32

33. Regression Statistics for BR Programs, Random effects 33 Type of ProgramnMStandard Error 95% Lower 95% Upper Z-value P- value Computer-assisted programs 330.280.060.160.404.710.000 Other BR programs470.390.040.320.479.840.000 Beginning Reading Total 800.350.030.290.4210.650.000

34. Regression Statistics for BR Programs, Random Effects 34

35. Meta-Analytic Multiple Regression Results from the Wilson/Lipsey SPSS Macro 35

36.  Investments in education have become an important national policy tool across the globe. With schools facing substantial costs of hardware and software, concerns naturally arise about the contribution of technology to students’ learning.  The present work lends some support to the proposition that computer-assisted interventions in reading are effective. The average effect for beginning reading computer-assisted programs is positive and substantively important (that is >0.25).  For the Beginning Reading topic area (grades K-3), the effect appears smaller than the effect achieved by non- computer reading programs. Conclusions 36

37.  Borenstein, M., Hedges, L.V., Higgins, J.P., and Rothstein, H.R. (2009). Introduction to meta-analysis. John Wiley and Sons.  Hedges, L. V. and Olkin I. (1985). Statistical Methods for Meta-Analysis. New York: Academic Press.  Lipsey, M.W., & Wilson, D.B. (2001). Practical meta-analysis. Thousand Oaks, CA: Sage.  Tobler, N.S., Roona, M.R., Ochshorn, P., Marshall, D.G., Streke, A.V., & Stackpole, K.M. (2000). School-based adolescent drug prevention programs: 1998 meta-analysis. Journal of Primary Prevention, 20(4), 275-336. References 37

38. Mathematica ® is a registered trademark of Mathematica Policy Research.  Please contact: –Andrei Streke AStreke@mathematica-mpr.com –Tsze Chan TChan@air.org For More Information 38

39. Mathematica ® is a registered trademark of Mathematica Policy Research. 39

40. Beginning Reading programs, random and fixed effects 40 Type of ProgramnM Standard Error 95% Lower 95% Upper Z-value P- value Computer-assisted programs 330.260.040.180.346.500.000 Other BR programs470.340.020.290.3914.350.000 Beginning Reading Total 800.320.020.280.3615.650.000 Type of ProgramnM Standard Error 95% Lower 95% Upper Z-value P- value Computer-assisted programs 330.280.060.160.404.710.000 Other BR programs470.390.040.320.479.840.000 Beginning Reading Total 800.350.030.290.4210.650.000

41. Computer-assisted programs, random and fixed effects 41

42.  Fixed effects model assume: (1) there is one true population effect that all studies are estimating (2) all of the variability between effect sizes is due to sampling error  Random effects model assume: (1) there are multiple (i.e., a distribution) of population effects that the studies are estimating (2) variability between effect sizes is due to sampling error + variability in the population of effects (Lipsey and Wilson, 2001) Random versus Fixed Effects Models 42

43. Beginning Reading Interventions, Random Effects, 95% Confidence Intervals 43

44.  Designs that confound study condition and study site –Programs that were tested with only one treatment and one control classroom or school  Non-comparable groups –Study designs that compared struggling readers to average or good readers to test a program’s effectiveness Examples of problematic study designs that do not meet WWC criteria 44