| Popis: |
The case-cohort study design is a cost-effective study design for large-scale follow-up studies, especially when the events are rare. Large-scale follow-up studies are also subject to competing risks – a situation where there are multiple causes of failure and the occurrence of one type of event prohibits the occurrence of the other types of events. There are two families of regression models in studying time to event data with competing risks: modeling for the cause-specific hazard or modeling the subdistribution (also called cumulative incidence function) of a competing risk. In this dissertation, we consider modeling the subdistribution of competing risk in case-cohort studies. We consider semiparametric proportional subdistribution hazards (PSH) and additive subdistribution hazards (ASH) models for an event of interest in the presence of competing risks in case-cohort studies. We propose estimating equations for each of the two models utilizing inverse probability weighting (IPW) techniques for the estimation of regression parameters. For both proposed methods of PSH and ASH models, the resulting estimators from their respective estimating equations are shown to be consistent and asymptotically normal under some regularity conditions. Further, simulation studies are carried out for both methods to examine the finite sample properties of the estimators. For illustration, the proposed methods are applied to case-cohort data from the Sister Study (Sandler et al. 2017). In this dissertation, we also considered the problem of testing the proportionality assumption in the PSH model with data from case-cohort studies. Schoenfeld-type residuals (Schoenfeld, 1982) from the estimating equation of the PSH model are formulated and, based on these residuals, correlation test and regression approach score test are proposed. Simulation studies demonstrate that the two proposed tests perform reasonably well for competing risks data in different case-cohort sample settings. Finally, the tests are applied to case-cohort data from the Sister Study. |
