Meta-Analysis: A Structural Equation Modeling Approach

5.7 Illustrations using R

This section illustrates how to conduct a multivariate meta-analysis using the metaSEM package. The first data set tests the effectiveness of the Bacillus Calmette–Guerin (BCG) vaccine for preventing tuberculosis (Colditz et al., 1994). This data set has been used by several researchers (e.g., Berkey et al., 1998; Viechtbauer, 2010). It is used to illustrate the multivariate random-effects meta-analysis. The second data set was extracted from the World Values Survey II (World Values Study Group, 1994). It has been used to demonstrate the SEM-based multivariate meta-analysis in Cheung (2013b). This data set is also used to illustrate the mediation and the moderation analyses. The statistics reported in the illustrations were captured by using the Sweave function in R. The numbers of decimal places may be slightly different for those reported in the selected output and in the text.

5.7.1 BCG vaccine for preventing tuberculosis

When we conduct a meta-analysis on contingency tables, we usually calculate the logarithm of the odds ratio between the treatment group and the control group as theeffect size. A univariate meta-analysis is applied to the data because there is only one effect size. van Houwelingen et al. (2002) argued that this practice may hide a lot of information, especially when there is a high degree of variation on the control and the intervention groups. They illustrated how a multivariate meta-analysis may be applied to the effect sizes ...