Time-dependent association for bivariate interval censored data

Abstract

Most statistical methods for the association between two failure times are global measures which have a shortage to mask a local dependency. To overcome this shortcoming, a cross ratio function can be considered as a local measure. Hu et al. (2011) suggested a polynomial function for log-transformed cross ratio function for bivariate right censored data and derived a pseudo-partial likelihood to estimate the parameters. In this paper, we extend this approach to bivariate interval censored data. Simulation studies are conducted to assess the finite sample properties of the proposed estimates. The methods are illustrated to investigate both the association pattern of dairy cow udder infection times and that of tooth eruption times and caries times from 500 children’s dental record. © 2021, Japanese Federation of Statistical Science Associations.