Showing papers by "Ralf Schwarzer published in 1988"

Meta-analysis programs

Abstract: The term meta-analysis (Glass, 1976) refers to a set of data analysis strategies for the integration and comparison of empirical research findings in the absence of the original data. A variety of advanced techniques for conducting meta-analyses are described in textbooks published recently (e.g., Hedges & Olkin, 1985; Rosenthal, 1984; Wolf, 1986). The present program package has been designed to provide a convenient tool for handling different kinds of data sets and different algorithms. It can be considered as a more advanced alternative to the programs offered by Gorman, Primavera, and Karras (1983). After typing in "META," a main menu appears with the following 10 subprogram options: 1. Data File Management. The ftler allows input of data according to three ftle structures, which are intended for use with either probabilities p, effect sizes d, or effect sizes r (correlations). The ftler allows the user to list, print, append, edit, and select data records. 2. Combination of EtTect Sizes d. This subroutine runs a meta-analysis for d values obtained from experimental studies. dis defined as the standardized difference between means, that is, the mean of the experimental group minus the mean of the control group or the pooled standard deviation (Fricke & Treinies, 1985; Glass, McGaw, & Smith, 1981; Hedges & Olkin, 1985). 3. Cluster Analysis of EtTect Sizes d. The distribution of effect sizes from a number of studies may be heterogeneous. Grouping smaller numbers of studies into more homogeneous subsets is achieved by a disjoint cluster analysis. By employing this strategy and by inspecting the resultant subgroups, potential moderator variables can be detected. 4. Combinations of EtTect Sizes r (Correlations). This program runs a meta-analysis for effect sizes r obtained either directly from correlational studies or from statistics that have been transformed with program 7. 5. Cluster Analysis ofEtTect Sizes r (Correlations). Same as in program 3 for effect sizes d. 6. Combination of Probabilities p. This program uses the Stouffer method of integrating one-tailed exact p values after transformation to the normal distribution Z. It should be used only when effect sizes are unavailable. 7. Transformation of Coefficients. Not all studies in the literature provide the appropriate effect sizes. Instead, some report t values, F values, chi-squares, or