Meta Analysis An Introduction. What… is… it? A “study of studies,” i.e., averaging results across studies in a given domain to get a better estimate of.

Slides:



Advertisements
Similar presentations
Power Analysis in Grant Writing Jill Harkavy-Friedman, Ph.D.
Advertisements

What is a sample? Epidemiology matters: a new introduction to methodological foundations Chapter 4.
Analysis of frequency counts with Chi square
Comparing the Various Types of Multiple Regression
Clustered or Multilevel Data
Lecture 24: Thurs. Dec. 4 Extra sum of squares F-tests (10.3) R-squared statistic (10.4.1) Residual plots (11.2) Influential observations (11.3,
PSY 1950 Confidence and Power December, Requisite Quote “The picturing of data allows us to be sensitive not only to the multiple hypotheses that.
Bivariate & Multivariate Regression correlation vs. prediction research prediction and relationship strength interpreting regression formulas process of.
Stat 112: Lecture 9 Notes Homework 3: Due next Thursday
Impact Evaluation Session VII Sampling and Power Jishnu Das November 2006.
Week 14 Chapter 16 – Partial Correlation and Multiple Regression and Correlation.
SW388R7 Data Analysis & Computers II Slide 1 Multiple Regression – Basic Relationships Purpose of multiple regression Different types of multiple regression.
Regression and Correlation Methods Judy Zhong Ph.D.
Issues in Experimental Design Reliability and ‘Error’
ANCOVA Lecture 9 Andrew Ainsworth. What is ANCOVA?
Overview of Meta-Analytic Data Analysis
Soc 3306a Lecture 8: Multivariate 1 Using Multiple Regression and Path Analysis to Model Causality.
ANOVA Greg C Elvers.
EVAL 6970: Cost Analysis for Evaluation Dr. Chris L. S. Coryn Nick Saxton Fall 2014.
Advanced Statistics for Researchers Meta-analysis and Systematic Review Avoiding bias in literature review and calculating effect sizes Dr. Chris Rakes.
RMTD 404 Lecture 8. 2 Power Recall what you learned about statistical errors in Chapter 4: Type I Error: Finding a difference when there is no true difference.
The Sample Variance © Chistine Crisp Edited by Dr Mike Hughes.
Statistical Power 1. First: Effect Size The size of the distance between two means in standardized units (not inferential). A measure of the impact of.
Soc 3306a Multiple Regression Testing a Model and Interpreting Coefficients.
Statistical Analysis Topic – Math skills requirements.
Organizational Psychology: A Scientist-Practitioner Approach Jex, S. M., & Britt, T. W. (2014) Prepared by: Christopher J. L. Cunningham, PhD University.
Chapter 9 Analyzing Data Multiple Variables. Basic Directions Review page 180 for basic directions on which way to proceed with your analysis Provides.
Funded through the ESRC’s Researcher Development Initiative Prof. Herb MarshMs. Alison O’MaraDr. Lars-Erik Malmberg Department of Education, University.
Multivariate Analysis. One-way ANOVA Tests the difference in the means of 2 or more nominal groups Tests the difference in the means of 2 or more nominal.
AP STATS: Take 10 minutes or so to complete your 7.1C quiz.
1 G Lect 7M Statistical power for regression Statistical interaction G Multiple Regression Week 7 (Monday)
The Campbell Collaborationwww.campbellcollaboration.org C2 Training: May 9 – 10, 2011 Introduction to meta-analysis.
Sampling, sample size estimation, and randomisation
Dependencies Complex Data in Meta-analysis. Common Dependencies Independent subgroups within a study (nested in lab?) Multiple outcomes on the same people.
Stat 112: Notes 2 Today’s class: Section 3.3. –Full description of simple linear regression model. –Checking the assumptions of the simple linear regression.
Analysis of Variance (One Factor). ANOVA Analysis of Variance Tests whether differences exist among population means categorized by only one factor or.
1 Lecture – Week 5 - Questionnaire Design & Selecting a Stats Test & Intro to G-Power. First - Some tidying Up According to my records there are a few.
ICCS 2009 IDB Workshop, 18 th February 2010, Madrid 1 Training Workshop on the ICCS 2009 database Weighting and Variance Estimation picture.
Stat 112 Notes 9 Today: –Multicollinearity (Chapter 4.6) –Multiple regression and causal inference.
Chapter 10 For Explaining Psychological Statistics, 4th ed. by B. Cohen 1 A perfect correlation implies the ability to predict one score from another perfectly.
KNR 445 Statistics t-tests Slide 1 Introduction to Hypothesis Testing The z-test.
LESSON 6: REGRESSION 2/21/12 EDUC 502: Introduction to Statistics.
D/RS 1013 Data Screening/Cleaning/ Preparation for Analyses.
Analysis of Experiments
Funded through the ESRC’s Researcher Development Initiative Department of Education, University of Oxford Session 2.1 – Revision of Day 1.
Multiple Regression David A. Kenny January 12, 2014.
Chapter 13 Understanding research results: statistical inference.
Biostatistics Regression and Correlation Methods Class #10 April 4, 2000.
Statistical Inference for the Mean Objectives: (Chapter 8&9, DeCoursey) -To understand the terms variance and standard error of a sample mean, Null Hypothesis,
Educational Research Inferential Statistics Chapter th Chapter 12- 8th Gay and Airasian.
June 25, Regional Educational Laboratory - Southwest Review of Evidence on the Effects of Teacher Professional Development on Student Achievement:
Meta-analysis Overview
Stats Methods at IC Lecture 3: Regression.
I-squared Conceptually, I-squared is the proportion of total variation due to ‘true’ differences between studies. Proportion of total variance due to.
Multiple Regression.
Analysis for Designs with Assignment of Both Clusters and Individuals
H676 Week 3 – Effect Sizes Additional week for coding?
Introduction to Regression Analysis
Meta-analysis: Conceptual and Methodological Introduction
RDI Meta-analysis workshop - Marsh, O'Mara, & Malmberg
Comparing several means: ANOVA (GLM 1)
Testing for moderators
Supplementary Table 1. PRISMA checklist
HLM with Educational Large-Scale Assessment Data: Restrictions on Inferences due to Limited Sample Sizes Sabine Meinck International Association.
CJT 765: Structural Equation Modeling
Week 14 Chapter 16 – Partial Correlation and Multiple Regression and Correlation.
Multiple Regression.
Spatial Data Analysis: Intro to Spatial Statistical Concepts
Incremental Partitioning of Variance (aka Hierarchical Regression)
UNDERSTANDING RESEARCH RESULTS: STATISTICAL INFERENCE
Presentation transcript:

Meta Analysis An Introduction

What… is… it? A “study of studies,” i.e., averaging results across studies in a given domain to get a better estimate of population parameters (Allen & Preiss, 2007; Hunter & Schmidt, 2004). Key ingredient = measures of effect size (usually Pearson’s r or Cohen’s d)

Why????????? 1. To reduce problems associated with SAMPLING ERROR at the individual study level. How big is this problem? According to Hunter, sampling is error is “massive” with sample sizes of N=100 or less, as we typically have in research. Result: VERY frequent Type II errors….

Example Monte Carlo study with population effect size (p) =.10, and 19 studies of sample sizes of N=30, N=68, and N=400. How many times did the conventional p <.05 test flag (*) an r as significant? 1/19 times, for a 95% error rate! What about for N = 68? 2/19, for an error rate of 89%. For N = 400, the error rate is STILL 47%!

Why????????? Hunter & Schmidt (1990) note that though Type I is typically 5%, Type II error rate for an average sample size (N = 80) with an average effect size (d =.40) and alpha level at p =.05 is… …still about 50%! How do we fix this? By increasing sample size in individual studies (rarely done) or through meta analysis.

But… it gets even worse…. In addition to random (sampling) error, studies also suffer from systematic problems, such as: – Measurement artifacts. – Issues of design. – Choice of sample. – Anything else that makes study results different. Fortunately, meta analysis can correct or at least account for these problems. Yay!

Points to Consider Meta analysis isn’t as easy as it may seem on the surface (see next slides), but… It provides the most accurate estimate of population parameters possible (vs. individual studies or literature reviews), and… FYI: Natural sciences also have variability in study results and most use forms of meta analysis to deal with divergent study findings.

So how can I do a meta analysis? It’s not available as an option in SPSS….  It seems like many meta analysts do the calculations by hand(!), though there is some software available for it, e.g., – Jack Hunter’s free DOS programs for “bare bones” meta analysis and other corrections he advocates for. – Commercially available software like CMA.CMA

Meta Analysis Example Sherry (2001) estimated the effect size of violent video games on aggression through meta analysis. Steps included: 1. Study selection and coding (p ) – Exhaustive lit search done for studies on video game violence and aggression, from – Of 900 initial returns, 25 studies were identified from which an effect size could be estimated. – Relevant info was recorded on coding sheets.

Meta Analysis Example 2. Effect sizes (r) estimated (p ) – Many had to be calculated from other stats. – Nonsignificant findings also had to be dealt with. 3. Overall effect size estimated (p ) – Mean effect size (weighted by sample size) and variance from all studies calculated. – Sampling error residual variance accounted for. – Variance from moderating variables accounted for. – See Table 1 (p. 420) for a typical summary.

Results Highlights Table 2 shows mean effect sizes and residual variance (p. 421) – High probability of moderating variables indicated. – Methodological variables include survey vs. experiment and type of outcome measure. – Theoretical variables include age, type of game violence, and length of game exposure. Table 3 shows how the theoretical variables correlate with effect size.

Results Highlights MULTIVARIATE STAT ALERT! These variables were entered into a multiple regression equation with effect size as the DV and moderators as IVs. Sherry’s reason? “To control for the effect of moderators on each other (e.g., suppression).” Check out the results in first full paragraph on p. 422.

What did we learn? Converting to d, overall effect size of games on aggression is.30, smaller than that found for television of.65 (Paik and Comstock, 1994). More recent games have a larger effect size. Player age also positively related to effect size. Effect size negatively related to playing time, however. Results used for theoretical advancement.

The Latest Meta Analysis Piece Theoretical assessment of evidence (2007): 1. No support for social learning theory. 2. Little support for catharsis, though not studied properly. Time finding points to it, as does declining violence at macro level. 3. Arousal (excitation transfer) and priming supported by available evidence. New model: PRIMED AROUSAL.

You Tube BREAK How prevalent is META ANALYSIS on YouTube? Not very…. Not very BONUS VIDEO tangentially related to meta analysis and the next slide: Powerthirst 2: Re- DominationPowerthirst 2: Re- Domination

Other Recent Examples Paul et al. (2007) Third Person Effect study— 32 studies (N = 45,729) indicate a substantial effect size of r =.50. Q: Moderators? A: Message, sampling, and respondent type. Other topics in Preiss et al. (2007) include effects of agenda setting, sexually explicit media, frightening media, music, health campaigns, spiral of silence, and more….

Final Considerations Shows importance of “replication, replication, REPLICATION” (Hunter 2001) in science. Some limitations, however…. Have to have a number of studies in a given domain before a meta analysis is worthwhile; otherwise, a literature review should suffice (Pfau, 2007). Q: Are there any areas you think are ripe for a meta analysis?

DO IT! Questions, comments, suggestions? Thank You.