Analysis of Multivariate Survival Times with Non-Proportional Hazards Models
In a clinical trial to evaluate treatments for a chronic disease, a commonly used regression method for analyzing multiple event times is based on a multivariate Cox model (Wei, Lin and Weissfeld, 1989). However, the Cox model may not fit the data well. For univariate survival analysis, a class of linear transformation models (Cheng, Wei and Ying, 1995a) provides many useful semi-parametric alternatives to the Cox model. In this paper, we take a similar approach as Wei et al. (1989) did for the multivariate case by modeling each marginal failure time with a linear transformation model and derive joint inference procedures for the regression parameters. In addition, we show how to check the adequacy of the fitted model graphically. We apply the proposed methods to data from an AIDS clinical trial and a cancer clinical trial for illustration.
KeywordsFailure Time Mycobacterium Avium Complex Recurrence Time Inference Procedure Proportional Odds Model
Unable to display preview. Download preview PDF.
- Byar, D. P. (1980). The Veterans Administration Study of Chemoprophylaxis for Recurrent Stage I Bladder Tumors: Comparisons of Placebo, Pyridoxine, and Topical Thiotepa. Bladder Tumors and Other Topics in Urological Oncology, eds Pavone-Macaluso, M., Smith, P. H. and Edsmyn, F., New York: Plenum, 363–370.Google Scholar
- Cheng, S. C., Wei, L. J. and Ying, Z. (1995b). Predicting survival probability with non-proportional hazards models, Technical Report, University of Texas, M. D. Anderson Cancer Center, Dept. of Biomathematics.Google Scholar
- Cheng, S. C., Wei, L. J. and Ying, Z. (1997). Predicting survival probability with semi-parametric transformation models. To appear in J. Am. Statist. Assoc.Google Scholar
- Fischl, M. A., Stanley, K. et al. (1995). Combination and monotherapy with zidovudine and zalcitabine in patients with advanced HIV disease. Annals of Internal Medicine 122, 24–32.Google Scholar
- Pettitt, A. N. (1982). Inference for the linear model using a likelihood based on ranks. J. R. Statist. Soc. C 33, 169–175.Google Scholar