LIKELIHOOD ANALYSIS FOR PROBIT REGRESSION WITH MEASUREMENT ERRORS

被引：9

作者：

SCHAFER, DW

机构：

[1] Department of Statistics, Oregon State University, Corvallis

来源：

BIOMETRIKA | 1993年 / 80卷 / 04期

基金：

美国国家科学基金会;

关键词：

EM ALGORITHM; ERRORS-IN-VARIABLES; GENERALIZED LINEAR MODEL; MEASUREMENT ERROR MODEL; STRUCTURAL MODEL;

D O I：

10.1093/biomet/80.4.899

中图分类号：

Q [生物科学];

学科分类号：

07 ; 0710 ; 09 ;

摘要：

Likelihood analysis is proposed for a probit regression model when one of the explanatory variables is measured with error and replicate measurements of that variable are available on some of the subjects. The distribution of the measurement error and of the unknown explanatory variable, conditional on the known variables, are both taken to be normal. The maximum likelihood estimates can be computed exactly with the EM algorithm. Likelihood ratio tests and confidence intervals can be computed with a Laplace approximation.

引用

页码：899 / 904

页数：6

共 6 条

[1]

BARNDORFFNIELSE.OE, 1989, ASYMPTOTIC TECHNIQUE

[2]

CARROLL RJ, 1984, BIOMETRIKA, V70, P19

[3]

CLAYTON DG, 1991, STAT MODELS LONGITUD, P301

[4] MAXIMUM LIKELIHOOD FROM INCOMPLETE DATA VIA EM ALGORITHM [J].

DEMPSTER, AP ;

LAIRD, NM ;

RUBIN, DB .

JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES B-METHODOLOGICAL, 1977, 39 (01) :1-38

[5] HETEROGENEITY IN MANTEL-HAENSZEL-TYPE MODELS [J].

LIU, Q ;

PIERCE, DA .

BIOMETRIKA, 1993, 80 (03) :543-556

[6]

SCHAFER DW, 1987, BIOMETRIKA, V74, P385, DOI 10.1093/biomet/74.2.385

← 1 →