On power analyses for individual site impacts in multisite trials

Slides presented at Society for Research on Education Effectiveness Conference 2022

Abstract: When conducting multisite trials, stakeholders often want precise estimates of site-level treatment effects. This paper provides advice for powering multisite trials to achieve this goal. We clarify common misconceptions about the properties of interval estimates in multisite trials and propose the use of an average margin-of-error metric for power analyses in these settings. We support our suggestions with an extensive simulation study and provide an example power analysis for the Scaling Up College Completion Efforts for Student Success (SUCCESS) initiative.