Abstract:
|
Beta regression allows for the modeling of outcomes that assume values in the open unit interval (0, 1). When the outcome assumes values in the closed unit interval [0, 1], then zero-one inflated beta regression models can be used. The current literature on beta regression and zero-one inflated beta regression includes approaches in both frequentist and Bayesian settings, but there have been few direct comparisons between the two approaches. In this talk, we conduct extensive simulation studies to compare the inferences between the frequentist and Bayesian approaches when 1) [0,1]-bounded outcomes have various probability mass at 0 and 1; 2) there is a single or multiple outcome variables; and 3) when outcome variables have repeated measures and are clustered. Recommendations will be made based on the pros and cons of the general methodologies and the simulation results.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.