Meta-analysis is the name of a methodology that encompasses the collection and analysis of independent studies. It has become an important and timely tool in the medical sciences. However the pace of technical (statistical) developments has now far surpassed the technical training of the research scientists who may wish to employ this promising new technique. Moreover, the technical literature is scattered over many journals in the statistical, medical, and social sciences. Readers approaching this literature face the difficulties that the technical level of many expositions is quite high and the conceptualization of models and the terminology used to describe them is quite diverse. We propose a monograph that would systematically unify and synthesize the literature on meta-analysis relevant to the medical sciences. It would draw on the emerging literature on meta-analysis in the social and behavioral sciences that we have already synthesized in earlier works. (See Statistical Methods in Metaanalysis by L.V. Hedges and I. Olkin, and Data Analysis for Clinical Medicine ed. by T.C. Chalmers.) To make this work useful to both technical specialists and to medical scientists, we plan to develop a manual that could be used by a scientist with a minimum of statistical training. Clearly, this will involve extensive computational development that can be used to perform both basic analyses (e.g., comparison of risks) and more advanced analyses that have been developed recently (e.g., nonparametric estimation of the distribution of random effects). The challenge will be to maintain the high degree of conceptual and technical excellence established by the book by Hedges and Olkin, while at the same time making the essential techniques reasonably understandable to the user without a formal education in statistics. This will be accomplished by developing innovative software to accompany the monograph and by undertaking constant interchanges between the statisticians on the one hand and the clinician scientists on the other.
