A Semiparametric Bootstrap for Proportional Hazards Models
We present a bootstrap resampling plan for the Cox partial likelihood estimator for proportional hazards models with nonrandom explanatory variables. Instead of resampling observed times, the proposed plan resamples from the Uniform(0,1) distribution of probability integral transformations of conditional failure times. The analysis can be completed without transforming resainpled values back into the original time scale, because the partial likelihood is invariant to monotone increasing transformations of the failure times. Adaptations to a variety of censoring schemes are discussed. A simulation study provides comparisons with standard partial likelihood estimation procedures and resampling plans that assume random explanatory variables.
KeywordsFailure Time Bootstrap Sample Partial Likelihood Product Limit Estimator Censoring Scheme
Unable to display preview. Download preview PDF.
- Andersen, P. K., Borgan, O., Gill, R. D. and Keiding, N. (1992), Statistical Models Based on Counting Processes, Springer-Verlag, New York.Google Scholar
- Breslow, N. (1974), “Covariance Analysis of Censored Survival Data,” Biometrics, 30, 89–99. Bryson, M. C. and Johnson, M. E. (1981), “The Incidence of Monotone Likelihood in the Cox Model,” Technonietrics, 23, 381–383.Google Scholar
- Loughin, T. M. (1993). Bootstrap Applications in Proportional Hazards Models, Ph.D. Dissertation, Iowa State University.Google Scholar