半参数变系数部分线性模型的函系数检验  

Tests for functional-coefficient parts on semiparametric varying-coefficient partially linear models

作　　者：黄振生[1] 

机构地区：[1]华东师范大学统计与精算学系,上海200241

出　　处：《华东师范大学学报（自然科学版）》2009年第4期54-61,共8页Journal of East China Normal University(Natural Science)

基　　金：华东师范大学2007优秀博士培养基金

摘　　要：通过局部线性方法给出未知函系数与未知参数的拟似然估计,利用未知量的拟似然估计构造了函系数部分的广义似然比检验,并证明了原假设下检验统计量渐近服从自由度独立讨厌参数与讨厌函数的x^2-分布,揭示了Wilks现象.通过模拟例子对给出的检验方法作出分析.Based on the quasi-likelihood estimates obtained by the local linear method, the generalized likelihood ratio tests for the models were established. Under the null hypothesis the test statistic followed asymptotically the x2-distribution with the scale constants and the degree of freedom being independent of the nuisance parameters and the nuisance functions, which was called the Wilks phenomenon. A simulated example was presented to illustrate the proposed methods.

关 键 词：局部线性方法 拟似然估计 广义似然比检验 Wilks现象 

分 类 号：O212.7[理学—概率论与数理统计]