During the past three decades, behavioral scientists have increasingly come to rely on methods of meta-analysis to provide quantitative summaries of the research literature. There is probably no area of organizational research where meta-analysis has not had a major impact. Consequently, a solid understanding of the statistical methods used in meta-analysis is important not only for those who apply the technique but also for researchers and practitioners who need to critically evaluate the many meta-analyses appearing in the research literature. Methods of Meta-Analysis: Correcting Error and Bias in Research Findings, by Hunter and Schmidt, is an excellent resource for learning about the rationale, computational procedures, and practical challenges faced when conducting a meta-analysis. This book is an update of the classic text by Hunter and Schmidt (1990), which introduced many organizational researchers to the techniques of meta-analysis. The new edition retains the features that made the original popular: a thorough explanation of the conceptual basis and computational procedures of meta-analysis, extensive coverage of methods for correcting statistical artifacts, and a style that is accessible to a wide audience. The methods presented in this book are the results of 30 years of vigorous research and debate and have been shown to provide accurate results under a wide variety of conditions. The new edition reflects several statistical refinements that have occurred since the publication of the 1990 text. In particular, there is extensive coverage of recently developed corrections for indirect range restriction. These new methods are shown to produce larger corrections than previous approaches and are likely to stimulate considerable interest among both proponents and critics of meta-analysis. Another change in the new edition is an increased emphasis on random-effect models. Many meta-analyses in the literature have used fixed-effect models, which assume a single population effect size is common to all studies. The more realistic random-effect model allows the population effect to differ for each study, requiring that researchers estimate not only the mean effect size but also the variance. Like the earlier edition, this book provides a thorough coverage of technical and practical issues in meta-analysis in a style that will be accessible to a wide audience. The authors make an effort to explain the methods without being overly mathematical, yet they still provide sufficient technical detail for the statistically oriented reader. In addition, numerous detailed examples help the reader to understand how to apply the technical details. For the most part, the book is successful in explaining the methods at a level appropriate for both novice and expert users of meta-analysis. However, the presentation of the technical details would be enhanced through the use of more consistent notation throughout the book. Subtle distinctions among the statistical terms can be difficult to grasp (e.g., distinguishing between the sampling variance of a single effect size, σej, and the variance across studies because of sampling error, σe), and this situation is not helped by changes in the notation. Similarly, I found the absence of notation to distinguish population parameters from their estimates to be confusing at points. Organizational Research Methods Volume 11 Number 1 January 2008 184-187 © 2008 Sage Publications 10.1177/1094428106295494 http://orm.sagepub.com hosted at http://online.sagepub.com
[1]
S. Raudenbush,et al.
A multivariate mixed linear model for meta-analysis.
,
1996
.
[2]
G. Barrie Wetherill,et al.
Random Effects Models
,
1981
.
[3]
Michael J. Strube,et al.
Validity Generalization: A Critical Review
,
2004
.
[4]
Ingram Olkin,et al.
Stochastically dependent effect sizes.
,
1994
.
[5]
Mark W. Lipsey,et al.
Practical Meta-Analysis
,
2000
.
[6]
Wolfgang Viechtbauer,et al.
Bias and Efficiency of Meta-Analytic Variance Estimators in the Random-Effects Model
,
2005
.
[7]
Diane M. Tomasic,et al.
Meta-Analysis: A Comparison of Approaches
,
2005
.
[8]
Frederick L. Oswald,et al.
On the robustness, bias, and stability of statistics from meta-analysis of correlation coefficients : Some initial Monte Carlo findings
,
1998
.