Shared Frailty Models Based on Cancer Data

TL;DR

This paper introduces new shared frailty models with inverse Gaussian and generalized Lindley distributions to better account for unobserved heterogeneity in cancer survival data, using Bayesian estimation.

Contribution

It proposes novel frailty models with specific distributions and applies Bayesian methods for parameter estimation in cancer data analysis.

Findings

Identified better-fitting frailty models for cancer datasets.

Demonstrated the effectiveness of Bayesian MCMC in parameter estimation.

Compared models using selection criteria to find optimal fits.

Abstract

Traditional survival analysis techniques focus on the occurrence of failures over the time. During analysis of such events, ignoring the related unobserved covariates or heterogeneity involved in data sample may leads us to adverse consequences. In this context, frailty models are the viable choice to investigate the effect of the unobserved covariates. In this article, we assume that frailty acts multiplicatively to hazard rate. We propose inverse Gaussian (IG) and generalized Lindley (GL) shared frailty models with generalized Weibull (GW) as baseline distribution in order to analyze the unobserved heterogeneity. To estimate the parameters in models, Bayesian paradigm of Markov Chain Monte Carlo technique has been proposed. Model selection criteria have been used for the comparison of models. Three different cancer data sets have been analyzed using the shared frailty models. Better models have been suggested for the data sets.