Department of Biostatistics Seminar/Workshop Series
Model assessment in multistate transition models through joint use of Bayesian variable selection and Bayes factors
David A. Engler, PhD
Assistant Professor, Department of Statistics, Brigham Young University
Wednesday, June 15, 1:30-2:30pm, MRBIII Conference Room 1220
In multiple sclerosis (MS), a primary clinical outcome measure is an ordinal score, the expanded disability severity score (EDSS). One of the main goals of MS research is to accurately model transitions between EDSS states over time. This objective first requires a number of significant modeling decisions, including selection of an appropriate link function and whether or not to collapse EDSS states. Then, under a selected model framework, it is also expected that the probability of transitioning from one state to another will vary across subjects; modeling will depend upon identification of transition-specific covariate effects. We propose joint use of Bayesian variable selection and Bayes factors in Markov transitional models to allow for both the selection of model framework and the identification of transition-specific covariate effects. Methods are assessed using both simulated data and data collected from the Partners MS Center in Boston, MA.