NONLINEARITY OF RANGE CORRECTIONS IN METAANALYSIS - TEST OF AN IMPROVED PROCEDURE

The authors evaluated an improved procedure for range-restriction corr ections in meta-analysis When population correlations were approximate ly normally distributed, the new nonlinear range-correction procedure improved the accuracy of the Schmidt-Hunter (S-H) interactive method i n estimating both the mean (M(rho)) and standard deviation (SD(rho)) o f population correlations, making it the most accurate of the procedur es examined. In the homogeneous case (SD(rho) = 0), the nonlinear rang e correction again improves accuracy of estimates of SD(rho). In this important case, Taylor Series Approximation (TSA) 1 is considerably le ss accurate than the S-H interactive and the TSA2 procedures, and N. S . Raju, M. J. Burke, J. Normand, and G. M. Langlois's (1991) procedure yields the least accurate estimates of SD(rho). Finally, the authors found that the nonlinear range-correction procedure produces improveme nts in the accuracy of the interactive method even under extreme viola tions of the normality assumption.