Power Struggles: Winning with Simulation-Based Sample Size Determination for Biomedical Research and Beyond

Thomas M. Menino Convention & Exhibition Center 

Room: CC-154 

This workshop shows participants how to perform simulation-based power analysis and sample size calculation. It is intended for statisticians, data scientists, and other data and data-savvy professionals involved in biomedical research or any other domain which uses statistically designed studies or experiments. The course includes a presentation of classical sample size approaches, as well as a discussion of their limitations even for relatively simple studies like one-factor experiments. Simulation is demonstrated as an attractive alternative which can be used for both simple and complex studies, in a way that precisely reflects the statistical analysis that will be done. We will focus on relatively standard models, such as ANOVA, logistic regression, and mixed models, but with an emphasis on the complexities of real-world statistical analysis and the flexibility of simulation-based power analysis to accommodate these complexities. Meta-issues, such as post-hoc power analysis, and working with collaborators to elicit necessary information will be discussed. Throughout the course, a wide variety of real examples along with active learning exercises will be provided. Participants will be equipped to give researchers a set of tradeoffs between resource efficiency and statistical efficiency, by building simulations which reflect the information available along with the statistical analysis plan.

Biometrics Section

Section on Statistical Consulting