多重二元响应Probit模型的渐近有效估计  

The Asymptotic Efficient Estimation of the Multi-binary Response Probit Model

作　　者：马建军[1] 徐兴忠[2] 

机构地区：[1]沈阳理工大学理学院,辽宁沈阳110168 [2]北京理工大学数学系,北京100081

出　　处：《生物数学学报》2008年第4期677-686,共10页Journal of Biomathematics

摘　　要：针对多重二元响应Probit模型提出了两步估计方法,第一步由边际似然得到参数n^(1/2)而相合的估计,第二步通过一步迭代得到渐近有效估计,由于只需一步迭代,因此在利用模拟方法计算信息阵时,可以增加模拟的次数,从而减少模拟所产生的扰动对估计的影响.A two-step method is proposed to estimate the parameters in the Multi-binary response probit model.The √n-consistent estimation can be derived from marginal maximum likelihood in the first step, and then an asymptotic efficient estimation can be derived by one-step iteration in the second step.The simulation method is employed to estimate the Fisher information matrix in the second step. Due to the only one-step iteration, the random number of the simulation integral can be increased to diminish the error of simulation.

关 键 词：LIMITED DEPENDENT variable模型 多重二元响应模型 PROBIT模型 极大似然估计 渐近有效估计 

分 类 号：O212.1[理学—概率论与数理统计]