London School of Hygiene & Tropical Medicine
About the Webinar
In this webinar, Dr. Jennifer Thompson discusses why you might consider using randomization tests to analyze data from complex trial designs such as stepped wedge trials. Randomization tests use the knowledge of the randomization scheme and the premise that if the intervention had no effect, it would not matter if or when clusters received the intervention; people would expect a similar result to the one observed. This allows you to calculate a p-value and construct confidence intervals for an intervention effect. Research has shown that this approach works well even with very few clusters and regardless of complexities like the correlation structure within clusters. The user-written command “swpermute” makes this approach easy to implement in Stata.
In this talk, Dr. Thompson describes what a randomization test is, discusses the benefits and drawbacks of this method, gives some real-world examples of when it can be most useful, and shows how you can implement this method in your own trials.
About Jennifer Thompson
Dr. Jennifer Thompson is an Assistant Professor in the International Statistics and Epidemiology Group at the London School of Hygiene & Tropical Medicine. She is co-leader of the methodology theme within her group as well as the quantifying impact theme in the Centre for Evaluation. Her expertise is in developing and evaluating analysis methods for stepped wedge and parallel cluster randomized trials and identifying more efficient trial designs. She has also been the statistician on several infectious disease treatment trials in low- and middle-income countries.