Studies on statistical analysis process of lifetime information

点击翻译按钮获取中文摘要

The purpose of this project is to develop statistical tools and methods which are useful to analyze data given in the medical fields, especially, useful to analyze lifetime (survival time or failure time) data and its related fields. If we interpret the lifetime as the time certain event occurs, then we may have data with the same structure in many fields of Reliability, Education, Psychology, Test Theory, Statistical Finance and so on. Hence the methods of lifetime analysis have enormous related fields. Lifetime data is taken by observing every living thing including human beings, animals, fishes and sso on, and due to their large variety the data is widely spread and skewed. Hence it is very difficult to apply classical parametric procedures where the normality of error distribution, the linearity of the mean structure and the homogeneity of error variance are assumed. Thus our attentions are focused on nonparametric procedures, semi-parametric methods, robust statistics and statisti … More cal graphics. Most of them are computer intensive methods.Head investigator and investigators studied statistical theory, methods and applications of the above procedures. Especially, we published papers and presented at several research meetings on (1)analysis of variance method for longitudinal data, (2)robust estimator based on the notion of depth, (3)decision of Wavelet bases functions by AIC, (4)diagnosis of the homogeneity of variances by spline smoothing, (5)construction of regression models by decision trees, (6)investigation of the properties of power transformations which are proposed to get normality, linearity and homogeneity of variances, (7)estimation of the causal effects, (8)developing computer intensive methods to analyze over-dispersed categorical data, (9)applications of the above procedures.Furthermore, we had scientific meetings five times. We had many participants including data analysts of private enterprizes. And we had exciting discussion useful both for research workers and data analysts. Less