Abstract: Sequentially observed survival times are of interest in many studies. For example, onemay be interested in time to recurrence and subsequent time to death following initialtreatment for cancer. There are difficulties in analyzing such data using nonparametricor semiparametric methods. First, when the duration of followup is limited and the timesfor a given individual are not independent, induced dependent censoring arises for thesecond and subsequent survival times. Non-identifiability of the marginal survivaldistributions for second and later times is another issue, since they are observable onlyif preceding survival times for an individual are uncensored. In addition, in somestudies a significant proportion of individuals may never have the first event. Fullyparametric models can deal with these features, but robustness is a concern. Weintroduce a new approach to address these issues. We model the joint distribution ofthe successive survival times by using copula functions, and provide semiparametricestimation procedures in which copula parameters are estimated without parametricassumptions on the marginal distributions. This provides more robust estimates andchecks on the fit of parametric models. The methodology is applied to a motivatingexample involving relapse and survival following colon cancer treatment.
Finally, this talk will discuss statistical methods for modeling complex time-to-eventphenotypes in genetic association studies, and current issues in identifying novelgenes involved in cancer recurrence following breast cancer treatment.
Date: Thursday, February 14, 2013
Time: 11:00 a.m.