Our research is diverse and most generally focuses on clustered data settings, where we can no longer rely on the typical assumption that all observations are independent of one another.
In cluster-randomized trials, rather than assign each person to a treatment, and the next person to a different treatment, we assign all the people in an administrative unit to the same treatment; it is the administrative units that are randomized. This gives rise to many interesting problems in planning and analysis.
We are developing software to calculate power for cluster-randomized trials. Power calculation is often treated as a post-hoc necessity for grant submission, but is actually an ethically necessary step in planning a trial. Grants from NIGMS (National Institute of General Medical Sciences R01GM121370) and UMass support this work.
We have a grant from the NIH ECHO (Environmental influences on Child Health Outcomes) program on issues around repeated measurements-- another instance of clustered data. In this project we examine the co-evolution of obesity and asthma and the multiple and intertwined exposures that cause these diseases.
See our articles page for recent developments in these areas.
STATISTICAL SCIENCE FOR PUBLIC HEALTH
Kleinman Lab at UMass Amherst