A Monte Carlo EM method for estimating multinomial probit models

被引：17

作者：

Natarajan, R

McCulloch, CE

Kiefer, NM

机构：

[1] Univ Florida, Dept Stat, Gainesville, FL 32610 USA

[2] Univ Florida, Div Biostat, Gainesville, FL 32610 USA

[3] Cornell Univ, Dept Stat Sci, Ithaca, NY 14853 USA

[4] Cornell Univ, Biometr Unit, Ithaca, NY 14853 USA

[5] Cornell Univ, Dept Econ, Ithaca, NY 14853 USA

来源：

COMPUTATIONAL STATISTICS & DATA ANALYSIS | 2000年 / 34卷 / 01期

基金：

美国国家科学基金会;

关键词：

Gibbs sampling; maximum likelihood estimation; menu pricing; Monte Carlo EM; observed information; panel data;

D O I：

10.1016/S0167-9473(99)00073-0

中图分类号：

TP39 [计算机的应用];

学科分类号：

081203 ; 0835 ;

摘要：

We develop a framework to perform maximum likelihood estimation in the multinomial probit model using a Monte Carlo EM algorithm. Our method includes a Gibbs step. This approach is different from likelihood procedures currently in use for this class of models, in that it does not involve direct evaluation and maximization of the observed data likelihood. Instead, we take advantage of the underlying continuum to simplify calculations. We also develop extensions of this Monte Carlo EM method for analyzing multi-period data. The computations are illustrated through real and simulated data. (C) 2000 Elsevier Science B.V. All rights reserved.

引用

页码：33 / 50

页数：18

共 30 条

[1] BAYESIAN-ANALYSIS OF BINARY AND POLYCHOTOMOUS RESPONSE DATA [J].