Summary: | 碩士 === 淡江大學 === 數學學系碩士班 === 101 === Regression diagnostic problems have been extensively studied for complete or right-censored survival data but so less for current status survival data. Here the current status survival data include an examination time, an status indicator for whether or not the failure has occurred by the examination time, and covariates. In this thesis, we have established four methods for evaluating the proportional odds assumption of current status survival data. They are the log odds curve plots, the observed and expected survival curve plots, the Cox-Snell residual method, and the Brier-score method. Also, the corresponding goodness-of-fit indices for four methods are proposed. Simulation results reveal good performance of the proposed methods and three real examples illustrate the applications of the proposed methods.
|