Choice of Parametric Accelerated Life and Proportional Hazards Models for Survival Data: Asymptotic Results

We discuss the impact of misspecifying fully parametric proportional hazards and accelerated life models. For the uncensored case, misspecified accelerated life models give asymptotically unbiased estimates of covariate effect, but the shape and scale parameters depend on the misspecification. The covariate, shape and scale parameters differ in the censored case. Parametric proportional hazards models do not have a sound justification for general use: estimates from misspecified models can be very biased, and misleading results for the shape of the hazard function can arise. Misspecified survival functions are more biased at the extremes than the centre. Asymptotic and first order results are compared. If a model is misspecified, the size of Wald tests will be underestimated. Use of the sandwich estimator of standard error gives tests of the correct size, but misspecification leads to a loss of power. Accelerated life models are more robust to misspecification because of their log-linear form. In preliminary data analysis, practitioners should investigate proportional hazards and accelerated life models; software is readily available for several such models.

[1]  R. A. Leibler,et al.  On Information and Sufficiency , 1951 .

[2]  D. Cox.  Regression Models and Life-Tables , 1972 .

[3]  Ross L. Prentice,et al.  Discrimination among some parametric models , 1975 .

[4]  D. Cox.  Tests of Separate Families of Hypotheses , 1961 .

[5]  Patricia J. Solomon,et al.  Parameter Orthogonality in Mixed Regression Models for Survival Data , 1997 .

[6]  David R. Cox,et al.  Regression models and life tables (with discussion , 1972 .

[7]  Ker-Chau Li,et al.  Regression Analysis Under Link Violation , 1989 .

[8]  J. Kent.  Robust properties of likelihood ratio tests , 1982 .

[9]  J Norrie,et al.  Model inconsistency, illustrated by the Cox proportional hazards model. , 1995, Statistics in medicine.

[10]  J. Lawless,et al.  Consistency and efficiency of regression coefficient estimates in location-scale models , 1988 .

[11]  Thomas R. Fleming,et al.  Model misspecification in proportional hazards regression , 1995 .

[12]  G. Kitagawa,et al.  Akaike Information Criterion Statistics , 1988 .

[13]  R. Davies,et al.  Statistical modeling of creep rupture data , 1999 .

[14]  M Schumacher,et al.  Effects of covariate omission and categorization when analysing randomized trials with the Cox model. , 1997, Statistics in medicine.

[15]  H. White.  Maximum Likelihood Estimation of Misspecified Models , 1982 .

[16]  John P. Klein,et al.  Effects of model misspecification in estimating covariate effects in survival analysis for small sample sizes , 1996 .

[17]  John D. Kalbfleisch,et al.  Misspecified proportional hazard models , 1986 .

[18]  D. Collett.  Modelling survival data , 1994 .

[19]  Paul Gustafson,et al.  On measuring sensitivity to parametric model misspecification , 2001 .

[20]  David Collett.  Modelling Survival Data in Medical Research , 1994 .

[21]  P. Andersen,et al.  Survival analysis 1982-1991: the second decade of the proportional hazards regression model. , 1991, Statistics in medicine.

[22]  N. Keiding,et al.  The role of frailty models and accelerated failure time models in describing heterogeneity due to omitted covariates. , 1997, Statistics in medicine.

[23]  P. J. Solomon.  Amendments and Corrections: `Effect of Misspecification of Regression Models in the Analysis of Survival Data , 1986 .

[24]  Patricia J. Solomon,et al.  Effect of misspecification of regression models in the analysis of survival data , 1984 .