学术报告（Xuewen Lu 7.10）

摘要：
In survival analysis, several scenarios exist where multiple events of interest are encountered in medical research and clinical trials. Examples for this include competing risks data, semi-competing risks data and multivariate failure time data, among others, they are all called multivariate survival data. Many models exist to enable estimation and investigation of relationships between covariates and joint timing of the multiple outcomes, but methods for model selection and prediction of these models in high dimensions are lacking.  Moreover, in such settings researchers commonly analyze only a single or composite outcome, ignoring other outcomes in estimation and model selection can cause loss of valuable information.  In this talk, we will discuss variable selection when one faces multivariate survival outcomes and various types of failure time data, such as left-truncated and right censored data or interval-censored data. For such problems, we present and discuss some recently developed penalized procedures and their theoretical properties and computational algorithms.