Clustering estimates of effect magnitude from independent studies

Larry V. Hedges*, Ingram Olkin

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

28 Scopus citations


One method of combining results of a series of studies is to calculate the average of the estimates of effect magnitude obtained from each study. The average estimate of effect magnitude may be misleading, however, when all studies do not share a common effect-magnitude parameter. When the effect-magnitude parameters (correlation coefficients or standardized mean differences) are heterogeneous across studies, it is often desirable to cluster studies into groups that are homogeneous with respect to the effect-size parameter. The present paper presents 2 procedures for clustering correlation coefficients and standardized mean differences when each estimator is based on the same number of observations. One procedure yields disjoint clusters and the other yields possibly overlapping clusters. In each case a method for determining the statistical significance level of the clusterings is given. Preliminary tests of homogeneity of a set of correlations or standardized mean differences are also given. The accuracy of the significance levels when estimators are based on different sample sizes is also studied. (21 ref) (PsycINFO Database Record (c) 2006 APA, all rights reserved).

Original languageEnglish (US)
Pages (from-to)563-573
Number of pages11
JournalPsychological bulletin
Issue number3
StatePublished - May 1983


  • procedures for clustering correlation coefficients & standardized mean differences, estimation of effect magnitude

ASJC Scopus subject areas

  • Psychology(all)


Dive into the research topics of 'Clustering estimates of effect magnitude from independent studies'. Together they form a unique fingerprint.

Cite this