Monday, 17 March 2014

Meta Analysis

"The scandalous failure of scientists to cumulate scientifically," a 2006 talk by Sir Iain Chalmers, referred to the tendency of scientists to look at individual studies in isolation, rather than as part of a systematic review of the "body of evidence" on a given subject.  Meta-analysis provides the statistical methodology to aggregate multiple studies for a general conclusion.  Learn more in Dr. Michael Borenstein's and Dr. Hannah Rothstein's online course, “Meta Analysis,” at For more details please visit at

One study reports a half-second longer reaction time for drivers using a cell phone, while another reports no effect from using a cell phone.  Conclusions from additional studies lie in between.  How to resolve these differences?  When done right, Meta analysis can aggregate multiple studies for a general conclusion. 

Aim of the course:
Meta-Analysis refers to the statistical analyses that are used to synthesize summary data from a series of studies. If the effect size (or treatment effect) is consistent across all the studies in the synthesis, then the meta-analysis yields a combined effect that is more precise than any of the separate estimates, and also allows us to conclude that the effect is robust across the kinds of studies sampled. By contrast, if the effect size (or treatment effect) varies from one study to the next, the meta-analysis may allow us to identify the reason for the variation and report (for example) that the treatment is more effective in a particular kind of patient, or in a particular dose range.

In this course we will discuss the logic of meta-analysis and the way that it is being used in many fields, including medicine, education, social science, ecology, business, and others. Participants will learn how to conduct a meta-analysis (how to compute an effect size, compute summary effects, assess heterogeneity of effects, test for differences in effect size across subgroups, and more). We will also discuss various controversies in meta-analysis (such as the question of mixing apples and oranges, the criticism of garbage-in-garbage-out). We will also draw on recent headline-making analyses such as the Avandia meta-analysis.

Participants will get hands-on experience in performing analyses using Excel(tm) and also using Comprehensive Meta-Analysis (CMA). All participants will have access to a free trial of CMA for the duration of the course. At the conclusion of the course, all participants should feel comfortable conducting a meta-analysis from start to finish using this or other software.

Who Should Take This Course:
Researchers who plan to perform a meta-analysis, or who want to be able understand meta-analyses that have been published by others.

Course Program:
·         Readings in the course text (see "Requirements" section)
·         Discussion forum with instructor
·         Homework (see below)
·         End of course data modeling project (see below)
·         Short narrated software demos
·         Supplemental readings available online
·         Supplemental - Archive of prior course discussions

Course outline: The course is structured as follows

SESSION 1: Computing the Overall Effect in Meta Analysis
  • What is meta analysis
o    Meta analysis in various fields
o    Meta analysis in medicine: Saving heart attack patients
o    Meta analysis in education: Some examples
o    Meta analysis in criminal justice: The "Scared Straight" jail program
  • The role of meta analysis
o    In planning research
o    In setting policy
  • Organizations for evidence-based policy
o    The Cochrane Collaboration (medicine)
o    The Campbell Collaboration (social science)
  • Computing a treatment effect
o    Focusing on treatment effects rather than p-values
o    From binary data
o    From continuous data
o    From correlational data
  • Computing an overall effect
o    Weighted means
o    Basic statistics
  • Forest plots
o    Basic issues

SESSION 2: Fixed vs. Random Effects in Meta Analysis
  • Heterogeneity among effect sizes
o    Assessing heterogeneity
  • Fixed effect vs. random effects models
o    Conceptual differences between these models
o    Computational formulas for these models

SESSION 3: Differences in Treatment Effects in Meta Analysis
  • Understanding differences in treatment effects
o    Moderator variables
o    Analysis of variance
o    Meta regression
  • Forest Plot
o    Advanced issues

SESSION 4: Publication Bias and other Issues in Meta Analysis
  • Publication bias
o    Funnel plots
  • Multiple subgroups within studies
  • Multiple outcomes within studies
  • Common criticisms of meta analysis
o    Apples and Oranges
o    Garbage in, Garbage out
o    Discrepancies between randomized trials and meta analyses

Homework in this course consists of short answer questions to test concepts, guided data analysis problems using software and end of course data modeling project.

In addition to assigned readings, this course also has an end of course data modeling project, short narrated software demos, and supplemental readings available online.

Dr. Michael Borenstein (Director, Biostat, Inc.) and Hannah R. Rothstein (Professor, Baruch College and the Graduate Center of the City University of New York) are co-authors (with Hedges and Higgins) of “Introduction to Meta-Analysis,” and co-editors (with Sutton) of “Publication Bias in Meta-Analysis: Prevention, Assessment and Adjustments.”

You will be able to ask questions and exchange comments with the instructors via a private discussion board throughout the course.   The courses take place online at in a series of 4 weekly lessons and assignments, and require about 15 hours/week.  Participate at your own convenience; there are no set times when you must be online. You have the flexibility to work a bit every day, if that is your preference, or concentrate your work in just a couple of days.

We, the Center for eLearning and Training (C-eLT), Pune, partner with and offer these courses to Indian participants at special prices payable in INR.

For India Registration and pricing, please visit us at

Call: 020 6680 0322


No comments: