Regression Analysis of the Generalized Accelerated Hazards Models With Interval‐Censored Data Under Case‐Cohort Studies

Abstract

Case‐cohort studies are typically undertaken in scenarios where disease incidence is low, or data collection on certain variables is costly. Case‐cohort researches with interval‐censored data often rely on survival models like the proportional hazards models, proportional odds models, accelerated failure time models and so on. However, these standard assumptions might not always be suitable for survival outcomes modelling in practical scenarios. To address this limitation, we provide a more adaptable class of generalized accelerated hazards models for analysing interval‐censored failure times in case‐cohort studies, for which there appears to be no well‐established approach in the literature. We develop a sieve maximum likelihood estimation method where the unknown cumulative baseline hazard function is integrated with regression parameters. The proposed estimators are proven to be consistent and normally distributed. A simulation study is carried out to assess the empirical performance of the proposed methods and suggest their practical effectiveness. Finally, we illustrate the approach using a dataset from a study on Alzheimer's disease.