
Title
A Bayesian Analysis of a Toxicity Study with Missing Data
Speaker
Mark J. Schervish, Carnegie Mellon University
Abstract
We use Bayesian hierarchical models to infer the relationships between the dose of a drinking water contaminant (perchlorate), blood levels of thyroid hormones and indicators of precancerous conditions in rats. Our statistical model is constructed to correspond to a mechanistic model for the effects of perchlorate on the response variables. The structure of our model helps us to deal with some issues that arise because certain useful data items were either not recorded or not collected.
This is joint work with Taeryon Choi, Ketra Schmitt and Mitchell Small.
Meet the speaker in Room 212 Cockins Hall at 4:30 p.m. Refreshments will be served.