DESIGNING EXPERIMENTS TO MAXIMIZE THE POWER OF DETECTING CORRELATIONS.

Studies investigating correlations among traits are increasingly common in evolutionary biology. By providing power calculations, minimum sample sizes, and standard errors of correlation coefficients in a variety of contexts, this note provides guidelines to insure that nonzero correlations will not be erroneously dismissed as nonsignificant in properly designed experiements. Forty individuals is often sufficient for reasonable estimation of correlation coefficients, although 100 or more individuals may be nexessary if a large number of traits are involved or comparisons are to be made among the coefficients. © 1998 The Society for the Study of Evolution.