|
Abstract:
|
Censored multivariate failure time data arise in many application areas. These data include both multiple time-to-respons measurements on individual study subjects, and correlated responses among study subjects. This talk is concerned with marginal methods for relating multivariate failure rates to corresponding regression variables, that may be evolving over time. Frailty methods are not so well suited to marginal estimation problems, because marginal failure time distributions involve frailty distribution parameters. Copula methods, which specify a joint distribution for cumulative hazard variates having a standardized distribution avoid this problem, but tend to lack flexibility and are not well suited to the inclusion of time-varying covariates. Hence, in this talk we will discuss models and estimation procedures for marginal hazard rate regression models and complementary dependency functions regression models. In the bivariate failure time special case both cross ratio regression models and covariance rate regression models will be considered for dependency modeling. Dependency modeling possibilities for higher dimensions failure time regression data will also be considered.
|
