1999 年 20 巻 1 号 p. 93-107
An application of the information criterion EIC for analysis of survival data in cancer of the thoracic esophagus is proposed. We propose Linearly Interpolated Kaplan-Meier (LIKM) model for the estimation of survival probability, and compare its performance with those of Weibull, gamma and exponential parametric survival probability models based on the maximum likelihood method. EIC based on model-based bootstrap samples is applied to the LIKM model and the parametric survivor models. It is shown that EIC makes it possible to compare goodness of the estimate obtained by the LIKM model with those of the parametric models. It is also shown that comparison of estimates obtained by the survivors models fitted to strata of categories based on EIC is useful for searching factors affecting survival probabilities after operations of cancer.