The Statistical Diagnosis For The Partial Nonlinear Model Of Random Cencorship

This article extends the statistical diagnosis of the partial linear randomly censored model in Li (2009) to the partial nonlinear model of random censorship. While the statistical diagnosis of the partly linear randomly censored model in Zhu (2001) has been fully studied, the statistical diagnosis of the partly nonlinear model of random censorship has not yet appeared in other paper. This paper introduces a new method of data synthesis to convert the original model to a new model. We find out outliers in the original mode through the statistical diagnosis for new model, The main work is the proof of the equivalence of the data deletion model and the mean shift model in the converted model. We present generalized Cook distance and so on for outliers and strong influencers and the local influence analysis are considered.This paper is divided into three chapters:In the first chapter, the main content is the introduction about the significance of statistical diagnosis, status development of model in this paper and some prelimilaries.In the second chapter, the statistical diagnosis of the partly nonlinear model under random censorship is considered. The main work is the proof of the equivalence of the data deletion model and the mean shift model in the converted model.We extend the statistical diagnosis of the partial linear randomly censored model in Li (2009) to the partial nonlinear model of random censorship.Our results extend and improve some known ones.In the third chapter, the main content is the local influence analysis about semiparametric regression model under random censorship and fixed design. First, we give the general method of local influence analysis, then analyse the weighted disturbance model and analyse the response variable disturbance model at last. We extend the exsisting methods and conclusions.