A note on averaging correlations

Abstract

Researchers are often faced with attempting to estimate a correlation based on the aggregation of sample rs either from several samples or from repeated measures within a single sample. The usual approach is either to average the observed correlations or to average the Fisher’s z- transformed rs and to back-transform the average z value. This note describes a minimum-variance unbiased estimator for the situation that is superior to either approach and is also simple to compute. © 1990, The Psychonomic Soceity, Inc.. All rights reserved.