چکیده مقاله
Survival analysis frequently encounters challenges related to data censoring and the need to model covariate effects beyond the conditional mean Quantile regression offers a robust alternative by exploring the entire conditional distribution of survival times, while semiparametric methods provide flexibility in modeling complex, nonlinear relationships This manuscript develops a Bayesian semiparametric framework for quantile regression with censored survival data The proposed model leverages the Asymmetric Laplace Distribution ALD to construct a valid likelihood for quantile estimation and employs penalized B splines to capture nonlinear covariate effects smoothly We specify a likelihood based on the ALD for a given quantile of interest, accommodating right, left, and interval censoring through appropriate likelihood contributions Penalized B splines P splines are used to model the linear predictor as a flexible function of covariates The model is implemented in a Bayesian framework using Markov Chain Monte Carlo MCMC via the Stan probabilistic programming language, which allows for efficient sampling and full uncertainty quantification Simulation studies demonstrate that the proposed method accurately recovers both regression coefficients and smooth functions across various censoring mechanisms and sample sizes The ALD based likelihood provides valid inference for conditional quantiles, while the P spline penalty effectively prevents overfitting An application to a simulated firm survival dataset illustrates the method's practical utility in estimating covariate effects on different quantiles of the survival time distribution, such as the median and lower tail The integration of the ALD with penalized splines within a Bayesian framework offers a powerful and flexible approach to quantile regression for censored survival data This method provides a more comprehensive view of the survival process than traditional mean regression models and is well suited for complex data structures common in biomedical and econometric research
کلیدواژهها
نویسندگان
Nastaran Najkar
Department of Agricultural Economics, International Campus of Ferdowsi University of Mashhad, Mashhad, Iran
شیوه ارجاع
Najkar، Nastaran،1404،Bayesian Semiparametric Quantile Regression for Censored Survival Data Using Penalized B-Splines and the Asymmetric Laplace Distribution،بیست دومین کنفرانس بین المللی پژوهش های نوین در مدیریت، اقتصاد، حسابداری و بانکدار�
ارائهشده در
مجموعه مقالات بیست دومین کنفرانس بین المللی پژوهش های نوین در مدیریت، اقتصاد، حسابداری و بانکداری26 اسفند 1404