Biostatistics Seminar Series - Cheng Yong Tang, PhD

Title: Predictive Survival Analysis with Longitudinal Measurements and Missing DataAbstract: An important practical problem in survival analysis is to predict the time to a future event. Available information in this scenario commonly includes longitudinal measurements from the same subject. Since future information is unknown and may develop dynamically over time, such a predictive problem is challenging. We consider a predictive approach based on modeling the forward intensity function. The forward intensity function approach is appealing for being capable of intrinsically incorporating the future dynamics that affect the stochastic event occurrence. To handle the practical difficulty due to the missing data in the longitudinal measurements, and to accommodate observations at irregularly spaced time points, we propose a smoothed composite likelihood approach for estimating the forward intensity model. Our theoretical analysis establishes the validity of the predictive modeling approach and the smoothed composite likelihood method. Extensive simulations and real data analysis demonstrate the promising performance of the predictive approach. This is a joint work with Lili Zhu.