Johnstone B M, Leino E V, Motoyoshi M M, Temple M T, Fillmore K M, Hartka E
Institute for Health & Aging, University of California, San Francisco 94133-3203.
Br J Addict. 1991 Oct;86(10):1211-20. doi: 10.1111/j.1360-0443.1991.tb01701.x.
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.
描述了酒精相关纵向合作项目所采用的研究设计和方法。该项目设计满足了开发评估酒精研究中研究结果可重复性程序的迫切需求。研究计划的关键特征包括：对一般人群饮酒行为多项纵向研究的原始数据进行重新分析；由跨学科核心人员开发并实施所有数据分析的集中化；与项目中纳入研究的原始研究者合作制定研究计划并解释结果；以及使用荟萃分析的建模程序来量化影响各研究中效应估计分布的因素的相对贡献，包括方法学差异和总体水平变量。最后一部分描述了该项目用于荟萃分析的统计方法，包括效应量估计的计算和合并程序、用于效应大小的分类和连续建模程序以及随机效应模型。
