A0469
Title: Rank-based regression for doubly interval-censored data
Authors: Seohyeon Park - Korea University (Korea, South) [presenting]
Sangbum Choi - Korea University (Korea, South)
Zhezhen Jin - Columbia University (United States)
Wenbin Lu - North Carolina State University (United States)
Abstract: In many biomedical fields, especially in studies of disease progressions, two sequential events where both event times tend are frequently encountered to be interval-censored due to regular examinations. Such a structure is called doubly interval censoring (DIC), and the primary interest is the elapsed time between two consecutive events. A weighted rank regression approach is proposed for DIC data under the semiparametric accelerated failure time model. After transforming DIC data into simple interval-censored data where true elapsed times may lie, estimating procedures are developed with a gehan-type weight by gathering all comparable pairs of observed residuals from transformed data. Moreover, it is generalized with data-dependent weights and is extended to clustered DIC data where the cluster size is potentially informative. An efficient resampling technique for the variance estimation is considered. Asymptotic properties are established, and numerical studies are conducted to demonstrate finite sample performances. Finally, the method with dental data is illustrated from the Signal Tandmobiel study to examine the effect of covariates on time to caries of four permanent first molars.
