Change in the center of the distribution and in the individual scores: Relation with heteroskedastic pre- and post-test distributions.

BACKGROUND Although average-based effect size (ES) and percentage of individual changes (PIC) are quite different, they are not independent: larger ESs lead to higher PICs. However, this association has not been sufficiently explored. METHOD We analyzed this association based on data simulated in the context of a pre-post design, with and without control groups. We simulated various distributions, sample sizes, degrees of test-retest reliability, effect sizes, and different variances in pre- and post-test. RESULTS The PIC is closely associated with the ES across a wide variety of empirically frequent scenarios. In the "single group pre-post designs", the linear regression model shows R2 values above 0.90. In the "control group pre-post designs", the linear regression model shows R2 values above 0.80. These results were found even when the post-test variability differed from that of the pre-test, replicating, extending and generalizing the findings in previous studies. CONCLUSIONS (1) In the absence of information about the PIC, the ES may be used to estimate this percentage. (2) The PIC is useful in interpreting the meaning of ES measures.