Activity Number:
|
214
|
Type:
|
Contributed
|
Date/Time:
|
Tuesday, August 13, 2002 : 10:30 AM to 12:20 PM
|
Sponsor:
|
Biopharmaceutical Section*
|
Abstract - #300940 |
Title:
|
Mixed Poisson Regression Analysis for Frailty Models
|
Author(s):
|
Shibao Feng*+ and Robert Wolfe
|
Affiliation(s):
|
University of Michigan and University of Michigan
|
Address:
|
1420 Washington Ht., Ann Arbor, Michigan, 48109-2029, USA
|
Keywords:
|
Multivariate survival analysis ; Frailty models ; Penalized quasi-likelihood ; Mixed Poisson regression models
|
Abstract:
|
The likelihood functions of both parametric multivariate survival frailty models (with piecewise constant baseline hazard) and semi-parametric multivariate frailty models are shown to be proportional to the likelihood functions of a class of mixed Poisson regression models. For multivaraite lognormal frailty models, the penalized quasi-likelihood (PQL) method can be applied as the inference procedure for the mixed Poisson regression models, although it is asymptotically biased. Thus a rich variety of random effect structures can be modeled. Simulation studies show that mixed Poisson regression models using PQL methods for inference yield satisfactory estimates for both the fixed and random effects of the multivariate lognormal frailty models. The procedure is illustrated with a national kidney transplantation dataset.
|