两值响应模型中的光滑统计推断-概率论与数理统计专业论文.docxVIP

两值响应模型的光滑统计推断 两值响应模型的光滑统计推断 英文摘要 Abstract Binary response model is a kind of regression model which the dependent variable only takes two value．Binary response model has been applied to many fields such as Biology， Medicine．Economy and Social Sciences．W色11一known Logistic regression model and Probit model are parametric examples of binary response model．Recently，semiparametric methods attract more and more attention．It provides more flexible model—based tools for statistical modeling and data analysis．As for the statistical inference and the asymptotic of regression coefficient in binary response model，researchers have got lot fruitful results，however，the complexity of those results and the difficulty to explanation bring many statistician to work out easier method，Horowitz proposed smoothed maximum score function estimator and the asymptotic distribution of the estimator is normal．However)the variance in the asymptotic distribution has unknown density function and it is not easy to be estimated accurately，SO we propose random weighting method to solve the difficulty，by using Monto—carlo methods， we estimate th．e asymptotic variance and avoid estimating the redundance parameters．In our paper we also study change point problem based on the model and find the estimator of the change point is consistent． There are four chapters in this thesis．In Chapter 1，we simply introduce some impor— rant literature on binary response models，and introduce some estimating methods of the regression coefficients．and the asymptotic of the estimator．胎also simply introduce some theory about change point and main results． In Chapter 2，the statistical modeling of some important generalized linear models is introduced． In Chapter 3 and Chapter 4，we introduce our main results more detailed and give the proof of these results． In chapter 3，we discuss the random weighting