METAANALYSIS AND THE SOLOMON 4-GROUP DESIGN

A potential weakness of quantitative methods is the intrusive nature o f testing in general, and pretesting specifically. The Solomon four-gr oup design ameliorates this difficulty, but because of statistical iss ues there are few published examples. W. Braver and Braver (1988) sugg ested the use of meta-analytic Stouffer's Z to combine the data from a ll four groups. Sawilowsky and Markman (1990a, 1990b, 1990c, 1990d) an d Braver and W. Braver (1990a, 1990b) exchanged differing opinions on this approach. The present study is a Monte Carlo demonstration that t he experiment-wise error rate inflates nearly triple nominal alpha. Ho wever, when not conducted as conditional tests, traditional procedures are more powerful than this meta-analytic approach, despite the abili ty of Stouffer's Z to combine all available data into a single statist ic.