How to use simulation methods for power analyses and sample size calculations in cluster-randomised designs

Speaker(s):

Jan R. Boehnke, University of Dundee

Abstract:

Randomising units such as schools or health services with the goal to evaluate innovative programmes, interventions, or other changes to their usual practice has opened new possibilities to evaluate research questions in real-world settings These so-called 'cluster-randomised' trials have become a cornerstone when planning research with impact and are used in a wide range of sectors and disciplines. The session will briefly introduce this study design, and then focus on a technical aspect: how to design (frequentist) case-based simulation studies to determine required sample sizes to test pre-specified hypotheses in R. The presentation will draw on multiple practical examples from the presenter's experience and offer (i) a structured approach to designing such studies (incl. code examples) and (ii) point to existing resources for specific design types.