Bayesian regression approach for modeling response time in clinical trials.

Sundaram, N.; Venkatesan, P.

International Journal of Computer Mathematical Applications; 2012; 6; 65-78.

Abstract: Cox proportional hazard (PH) regression models are most popular and widely used for analysis of censored survival data. The accelerated failure time (AFT) model, which regresses the logarithm of the survival time over the covariats, has rarely been utilized in the analysis of censored survival data. It has been used extensively to examine the covariate with random effects on the hazard function. The AFT models can be used to describe the influence of unobserved heterogeneity (Frailty) in a non-parametric and parametric PH models. The Exponentiated Exponential (EE) model also known as Generalized Exponential (GE) model fits as an alternative to life time models. In this paper an attempt has been made to model the censored survival data using random effect and Gamma frailty regression models. The Bayesian regressions with Markov Chain Monte Carlo (MCMC) methods are also studied. The parametric regression models estimate the parameter more efficiently than the Cox regression model. Bayesian Log-Normal regression model are found to be providing better fit than the other Bayesian regression models namely Exponential, GE, Webull, Log-Logistic and Gamma. Also it has been seen that the Bayesian regression models gave better fit than the classical regression models.

Keywords: AFT model; PH model; Frailty model; Hazard function; Bayesian model; MCMC.

Back to List of publications / Home