A single estimate of separate within-individual correlations: applications in psychopharmacology

Simple procedures are proposed to describe relationships between sets of repeated measures within study subjects. An example is the association between blood levels of prolactin and a drug (e.g., methylphenidate) measured at multiple time-points across study subjects. These procedures consist of (a) estimating Pearson product moment correlation coefficients (rs) within individuals using all the observations available per individual; (b) estimating an average r that pools the information across study subjects; (c) testing whether these rs are homogenous using a simple chi-square test; and (d) testing the average r for statistical significance. If these individual rs are found to be homogeneous, the average r is taken as representative of the individual rs. Testing rs for homogeneity and pooling them across individuals are relatively old procedures (Snedecor 1956). Recent applications can be found in the meta-analytic literature (Rosenthal 1983). Applications are demonstrated using actual data. We recommend using these procedures because they are simple and easily applied and interpreted.