Title
Statistical methods for bivariate survival data with interval sampling and application to biomedical studies
Speaker
Dr. Hong Zhu, The Ohio State University
Abstract
In biomedical cohort studies, it is common to collect data with incidence of a disease occurring within a calendar time interval. Bivariate or multivariate survival data arise frequently in these studies and are of interest when theordered multiple failure events are considered as the major outcomes to identify the progression of a disease. We consider an interval sampling scheme, where the rst failure event (i.e., cancer onset) is identied within a calender time interval, the occurrence of the initiating event (i.e., birth) can be retrospectively conrmed, and the observation of the second failure event (i.e.,death) is subject to right censoring. To analyze this type of bivariate survival data, it is important to recognize the presence of bias arising due to interval sampling. Under stationary and semi-stationary assumptions, we develop nonparametric and semiparametric methods to estimate the joint survival function. Simulation studies demonstrate the proposed estimating methods perform well with practical sample sizes in dierent simulated models. We apply the methods of joint survival function estimation to NCI's SEER ovarian cancer registry data for illustration of the methods and theory. Moreover, it is important to incorporate sampling bias in analyzing this type of bivariate survival data based on both uncensored and censored observations. We extend the work by developing a copula models approach and studying the dependence structure of bivariate survival data under the model assumptions. The copula models method is evaluated by simulations and illustrated by Rakai Human Immunodeciency Virus (HIV) seroconversion data to study the disease progression of HIV infection for treatment-naive individuals.
