COMPARISON OF 3 MAJOR METAANALYTIC APPROACHES

Three major meta-analytic approaches have been developed and applied to integrate empirical research. A comparison of their statistical formulas suggested that they should yield different answers to the meta-analytic questions of central tendency, variability, and prediction by moderators. These frameworks were used to analyze systematically differing databases and showed that, although the techniques of L. V. Hedges and I. Olkin (1985)and of R. Rosenthal and D. Rubin(e.g., 1978, 1988; Rosenthal, 1991)tended to produce reasonable and convergent results, the results of J. E. Hunter, F. L. Schmidt, and G. B. Jackson( 1982; Hunter and Schmidt, 1990) often diverged from the other 2 frameworks. For example, consistent with the law of large numbers, finding more studies with the same result is less likely to occur because of chance alone; the Hedges and Olkin and the Rosenthal and Rubin approaches confirmed this prediction, but the Hunter et al. approach did not. The distinct tendency of the Hunter et al. framework to produce results that violate conventional expectations suggests that it should be used with caution.