Application Of Shared Gamma And Inverse-Gaussian Frailty Models To Cancer Data

Abstract

Shared Gamma and Inverse-Gaussian Frailty models are used to analyze the survival times of patients who are clustered according to cancer/tumor types under Parametric Proportional Hazard framework. The result of the analysis shows that patients within the same cluster share some common unobserved heterogeneity which proves the analysis of such data under the assumption of independence inappropriate. This is an indication that conclusion based only on the observed covariates, ignoring the characteristic unobservable covariates in the data can be misleading. However, no evidence is strong enough for preference of either Gamma or Inverse Gaussian Frailty.

Keywords: Failure time, Shared Frailty, Baseline hazard.

Global Journal of Mathematical Sciences Vol. 7 (1) 2008: pp. 35-48