An integrated approach to meta-analysis in alcohol studies.

The research design and methods utilized by the Collaborative Alcohol-Related Longitudinal Project are described. The project design addresses the critical need to develop procedures to assess the replicability of research results in alcohol studies. Key features of the research plan include: re-analysis of original data from multiple longitudinal studies of drinking behavior in the general population; centralization of all data analyses, developed and implemented by an interdisciplinary core staff; development of the research plan and interpretation of results in co-operation with original investigators of studies included in the project; and use of modeling procedures from meta-analysis to quantify the relative contribution of factors influencing the distribution of effect estimates across studies, including both methodological differences and aggregate level variables. The final section describes statistical methods for meta-analysis used by the project, including procedures for the calculation and combination of estimates of effect magnitude, categorical and continuous modeling procedures for use with effect sizes, and random effects models.