Fitting Semi-Parametric Frailty Models with Poisson-Distributed Random Effects

• 발행기관: 한국자료분석학회
• 저자명: Il Do Ha Maengsek Noh
• 간행물 정보: 『Journal of The Korean Data Analysis Society (JKDAS)』Vol.13 No.6, 2763~2769쪽, 전체 7쪽
• 주제분류: 자연과학 > 통계학
• 파일형태: PDF
• 발행일자: 2011.12.30

영문 초록

Frailty models with parametric or non-parametric baseline hazards have been widely used due to the consideration of heterogeneity or correlation in analysis of survival data. For the frailty distribution, a continuous distribution such as gamma-distributed frailty (or random effect) has been usually assumed. However, the frailty distribution is unobservable, and therefore the choice of frailty distribution should be based on general modelling considerations. Accordingly, a discrete-distributed frailty can be considered. In this paper, we propose a semi-parametric frailty model allowing non-parametric baseline hazard and Poisson distributed frailty. For the model fitting we present a profile likelihood-based estimation procedure. The proposed method is illustrated using a well-known practical survival data set. We find out that the Poisson frailty can be an useful alternative to non-PH survival data.

목차

1. Introduction

2. Frailty model with Poisson frailty

3. Estimation procedure

4. Practical example

5. Discussion

References

키워드

Frailty models Laplace transform Marginal likelihood Semi-parametric model Discrete frailty Random effect