ScholarWorks@숙명여자대학교

초록

Interval-censored competing risk data occurs when subjects are at risk of several causes and the failure time is known to lie at somewhere between two sequential observation times. For investigating the effect of covariate on the hazard, Goetghebeur and Ryan (2000)s approximate likelihood is extended with a piecewise baseline hazard assumption. The EM algorithm is implemented to update unknown event status and risk status and the IPCW technique is applied to adjust the effect of the subjects experiencing competing events. Simulation results show the estimates based on the suggested method are unbiased and satisfy the coverage probabilities. Two HIV-related datasets (HIV data from intcrr package and the Bangkok Metropolitan Administration (BMA) prospective study HIV-seronegative injecting drug users data) are analyzed as real examples.

키워드