B0378
Title: The weighted least squares method for heteroscedastic interval censored survival data
Authors: Najmeh Nakhaeirad - University of Pretoria (South Africa) [presenting]
Ding-Geng Chen - College of Health Solution- Arizona State University-Phoenix- USA (United States)
Abstract: In clinical and epidemiological research, the failure times are observed exactly or within certain intervals, such as in HIV infection, which is called as partly interval-censored data. If an individual takes frequent visits, then the occurrence of an asymptomatic event can be determined with sufficient accuracy, while when the visits are infrequent, the occurrence of an asymptomatic event is known to lie within an interval that may be too broad to be treated as exact. In the analysis of time-to-event data, many approaches have been proposed to estimate the parameter of the accelerated failure time model (AFT), which is popular due to its simplicity and ease of interpretability. In the classical AFT model, the random errors are assumed independent and identically distributed, and independent of the covariates despite the fact that in practice, the random errors depend on the covariates which exhibit heteroscedasticity. In this regard, the semi-parametric weighted least squares method is extended to accommodate heteroscedastic partly interval-censored survival data. A resampling method is developed to estimate the variance of the parameter estimates. A simulation study is conducted to assess the performance of the proposed approach in the presence of interval-censored data and to evaluate the resampling method. Finally, a real dataset is analyzed for illustration.
