Online Program

Effective Communication with Clients to Estimate Effect Size for Power Analysis (303002)

Keywords: effective client communication, power analysis, effect size, sample size, statistical significance, practical importance

Power analysis as a justification of a sample size for a designed study is typically the first-encountered major task in statistical practice. It is, however, often challenging when working with clients to obtain from them reasonable estimates of the required parameters. Statistical power is a function of three parameters: the alpha level (significance threshold), the sample size, and the effect size. Thus, knowledge of three values is required to solve for the fourth value. Typically, the alpha level is set at 0.05 and the statistical power at 80%. This leaves only estimation of the effect size to solve for the sample size. This presentation will review the various types of effect sizes (d, r, f2, and PV) and provide critiques on the use of current conventions (small, medium, and large) for defining effect size and on the existing gap between statistical significance and practical importance. Finally, I will demonstrate how to use a meta-analytic review, which classifies multidisciplinary outcome variables by effect size, to guide client communications to develop valid and accurate estimate of this final required parameter, effect size.