Meta-Analytic Procedures for Social Research

by Robert Rosenthal

Book Overview

Praised in the first edition for the clarity of his general framework for conceptualizing meta-analysis, Rosenthal′s revised edition covers the latest techniques in the field, such as a new effect size indicator for one size data, a new coefficient of robustness of replication, new procedures for combining and comparing effect sizes for multiple dependent variables, and new data on the magnitude of the problem of incomplete retrieval (the...

• Format: Paperback
• Language: English
• ISBN: 080394246X
• ISBN13: 9780803942462
• Release Date: May 1991
• Publisher: Sage Publications, Inc
• Length: 168 Pages
• Weight: 0.55 lbs.
• Dimensions: 0.4" x 5.5" x 8.6"

Customer Reviews

Meta-analytic procedures for social research

Quite a good book covering most important aspects of meta-analysis. I would criticize only one aspect. The one of mathematical rigor. The assumptions that should apply in order to use a technique are not always stated. That is the case with several other books on this topic. Also, I would argue against the use of z/sqrt(N) as an estimate for Pearson's r which is used throughout the book and is described as a 'good approximation', and as such propagated through the literature from here. One could easily check that the maximum of that estimate for Wilcoxon test with balanced samples converges towards 0.866... as the samples grow larger, which is far from 1. On the other hand, for some tests (like Gehan's test for censored data) it can be as large as 1.3. Hence, it is far from good approximation.

Lucid, concise, informative - a gem for any researcher!

Whether you are simply interested in an ability to understand meta-analytic studies or have already published such works, this small text will be of great assistance. The author provides clear and concise discussion of all the relevant issues including valuable formulas for obtaining effect sizes from various tests of significance, procedures for appropriately combining effect sizes and testing for significance, and an interesting discussion of the binomial effect size display. While such information and methods are available elsewhere, I have never seen such a clear and concise presentation. There is also an excellent discussion of the file-drawer problem and alternatives for addressing it. In short, this is an elegant work that belongs on the shelves of any life-long student of the social sciences!