|
The Handbook of Research Synthesis
|
|---|
Harris Cooper and Larry V. Hedges; eds. 573 pages. New York: Russell Sage Foundation; 1994. $49.95.
Meta-analysis has evolved in response to the need to synthesize the burgeoning literature for the advancement of knowledge. Not without irony, the literature on meta-analysis itself is threatening to reach the explosion limit. Do we need another book on meta-analysis? Many have focused on statistical techniques for meta-analysis, and, to its credit, this book offers a much broader spectrum of topics without sacrificing depth. For example, The Handbook of Research Synthesis integrates methods of literature search, management of data sets, and effective communication of results, as well as several statistical techniques, into its discussion.
Although both behavioral and medical scientists are the intended audience, the examples and terminology are taken more often from behavioral science. For example, meta-analysis is narrowly defined as the statistical analysis of results from several studies, whereas its accepted definition in medicine, right or wrong, is the entire process of synthesizing research. The book covers the entire process and labels it research synthesis, and this difference in semantics extends throughout the text. One advantage of the social sciences viewpoint, however, is that it offers solutions, absent in the medical literature, on how to integrate the qualitative and quantitative aspects of observational studies.
Intended for those who carry out research synthesis, this is by no means an introductory text. It is an eloquently written reference book of 32 chapters by 43 authors, plus a glossary, index, and appendixes containing data sets. As I see it, if you are doing a meta-analysis and wish to have in-depth knowledge on a certain topic, such as systematic coding for research synthesis, fixed- and random-effects models, or sensitivity analysis and diagnostics, this is the sourcebook for you. Issues here that I have not seen covered elsewhere in the literature on meta-analysis include ways to handle the lack of independence that arises in studies with multiple treatments or outcomes. The book thoroughly covers topics such as missing data and publication bias and introduces Bayesian approaches. You will not find information on software packages, pooling of primary data, or analysis of exposure-response relations.
Practitioners of meta-analysis in any field can certainly benefit from the main premise of this book: The same rigorous standards must be applied to research synthesis that are applied to the original research studies. Meta-analyses are easy to do but difficult to do well. This book is a welcome addition to the continual endeavor to improve the standards of conduct and reporting of research syntheses.
Article
