Meta-analysis: Its role in psychological methodology

Meta-analysis refers to the statistical analysis of a large collection of independent observations for the purpose of integrating results. The main objectives of this article are to define meta-analysis as a method of data integration, to draw attention to some particularities of its use, and to encourage researchers to use meta-analysis in their work. The benefits of meta-analysis include more effective exploitation of existing data from independent sources and contribution to more powerful domain knowledge. It may also serve as a support tool to generate new research hypothesis. The idea of combining results of independent studies addressing the same research question dates back to sixteenth century. Metaanalysis was reinvented in 1976 by Glass, to refute the conclusion of an eminent colleague, Eysenck, that psychotherapy was essentially ineffective. We review some major historical landmarks of metaanalysis and its statistical background. We present the concept of effect size measure, the problem of heterogeneity and two models which are used to combine individual effect sizes (fixed and random effect model) in great details. Two visualization techniques, forest and funnel plot graphics are demonstrated. We developed RMetaWeb, simple and fast web server application to conduct meta-analysis online. RMetaWeb is the first web meta-analysis application and is completely based on R software environment for statistical computing and graphics