Copula functions and bivariate distributions for （连系动词的功能和二元分布）.pdfVIP

Copula functions and bivariate distributions for survival analysis: An application to political survival Alejandro Quiroz Flores Wilf Department of Politics New York University 19 West 4th St, Second Floor New York, NY 10012-1119 aqf200@ July 20, 2008 Abstract Event-history analysis often focuses on the survival time of a single subject. However, recent research in the social sciences demands estimation of the joint survival time of different subjects. This paper presents a method to estimate the interdependence between two different subjects. Analogous to seemingly unrelated regressions (SUR) or bivariate probit models, this paper begins with the assumption that the two different survival processes are not independent. The interdependence between processes is modelled as part of a bivariate distribution suit- able for survival analysis, such as the bivariate exponential and the bivariate Weibull. These bivariate distributions are derived from copula functions. To test the performance of these dis- tributions, the paper presents a simulation experiment. In order to illustrate these methods, the paper presents an application to a new data set on the tenure of leaders and foreign ministers. 1 1 Introduction Event-history analysis often focuses on the survival time of a single subject. However, recent re- search in the social sciences demands estimation of the joint survival time of different subjects. This paper presents a method to estimate the interdependence between two different subjects. Ana