SeminarsCopula modeling for dependent truncation data- Review and new development
reads
Takeshi Emura
2011-03-11
12:45:00 - 14:45:00
405 , Mathematics Research Center Building (ori. New Math. Bldg.)
In recent years, copula modeling for multivariate random variables has received increasing attentions. Most existing methods for analyzing truncation data, including the Lynden-Bell estimator, critically rely on the assumption of the independence between lifetime and truncation time. In this talk, we introduce the concept of "copula" which is a flexible tool to describe the dependence between two random variables. To relax the assumption of independence between lifetime and truncation time, we also introduce a"semi-survival copula" modeling approach proposed by Chaieb et al. (2006, Biometrika), a suitable copula model for dependent truncation. Then, we discuss the proposed inference procedure using the nonparametric maximum likelihood estimator (NPMLE) under the semi-survival copula model. We compare the proposed method with existing methods. The talk is intended to give overall understanding in the current research on copula modeling for dependent truncation. (This is joint work with Weijing Wang of the Institute of Statistics, National Chiao Tung University.)