SPEAKER:              Giovanni Parmigiani
                      Institute of Statistics and 
                         Decision Sciences
                      Duke University

TITLE:                Bayesian Approaches to Survival Data
                      in Genetic Epidemiology
                      

                           ABSTRACT

     In this lecture I will review and illustrate Bayesian
approaches to survival data that arise from family studies of
genetic susceptibility to cancer. I will focus on three main
methodological themes: the error-in-measurement modeling
of unobserved genotypes; the modeling of the ascertainment
mechanism in family studies of high risk groups; and the
implementation of flexible semiparametric model
specification.

     Some specific issues will be illustrated using
ongoing studies about breast/ovarian cancer
susceptibility genes BRCA1 and BRCA2.  In particular I
will describe a technique for estimating effects associated
with genetic status in the Cox proportional hazards
model when the genetic status of members of the study
group is unknown.  The result is an a posteriori sample
of model parameters that accounts for sampling
error, uncertainty in the genetic status of study
participants, and uncertainty in estimates of population
parameters.

     Related materials are available at url
www.isds.duke.edu/~gp