Bayesian Regression Methods for Crossing Survival Curves

variational methods;Bayesian inference;survival analysis

DiCasoli, Carl Matthew ; Dr. Charles Apperson, Committee Member,Dr. Wenbin Lu, Committee Member,Dr. Brian Reich, Committee Member,Dr. Subhashis Ghosal, Committee Co-Chair,Dr. Sujit Ghosh, Committee Chair,DiCasoli, Carl Matthew ; Dr. Charles Apperson ; Committee Member ; Dr. Wenbin Lu ; Committee Member ; Dr. Brian Reich ; Committee Member ; Dr. Subhashis Ghosal ; Committee Co-Chair ; Dr. Sujit Ghosh ; Committee Chair

University:North Carolina State University

关键词: variational methods;    Bayesian inference;    survival analysis;   

Others  :  https://repository.lib.ncsu.edu/bitstream/handle/1840.16/4743/etd.pdf?sequence=1&isAllowed=y

美国|英语

来源: null

【 摘 要 】

In survival data analysis, the proportional hazards (PH), accelerated failure time(AFT), and proportional odds (PO) models are commonly used semiparametric models forthe comparison of survivability in subjects. These models assume that the survival curvesdo not cross. However, in some clinical applications, the survival curves pertaining to the two groups of subjects under the study may cross each other, especially for long-durationstudies. Hence, these three models stated above may no longer be suitable for making inferenceYang and Prentice (2005) proposed a model which separately models the short-term and long-term hazard ratios nesting both PH and PO. This feature allows for the survival functions to cross. First, we study the estimation procedure in the Yang-Prentice model with regards to the two-sample case. We propose two different approaches: (1) Bayesian bootstrap and (2) smoothing methods. The first approach involves Bayesian bootstrapwith likelihoods corresponding to binomial and Poisson forms while the second approachinvolves kernel smoothing methods as well as smoothing spline methods. A simulation isconducted to compare various methods under the two-sample case. Next, we extend theYang-Prentice model to a regression version involving predictors and examine three likelihoodapproaches including Poisson form, pseudo-likelihood, and Bayesian smoothing. Theeffects of model misspecification on asymptotic relative efficiency are also studied empirically. The results from simulation studies indicate that the PH, AFT, and PO models are not robust to model misspecifications when the survival functions are allowed to cross.Finally, we calculate the marginal density via variational methods to determinethe Bayes factor. Either a full Bayesian or Bayesian approach is implemented to performmodel selection. Both approaches accurately identify the correct model, even under slightmisspecification, and are computationally more efficient than MCMC techniques.