Maximum likelihood estimation of a mixture of normal regressions: Starting values and singularities

被引：13

作者：

Caudill, SB ^{[1
]}

Acharya, RN ^{[1
]}

机构：

[1] Auburn Univ, Dept Econ, Auburn, AL 36849 USA

来源：

COMMUNICATIONS IN STATISTICS-SIMULATION AND COMPUTATION | 1998年 / 27卷 / 03期

关键词：

mixture model; convergence; singularities; starting values;

D O I：

10.1080/03610919808813502

中图分类号：

O21 [概率论与数理统计]; C8 [统计学];

学科分类号：

020208 ; 070103 ; 0714 ;

摘要：

A Monte Carlo simulation was conducted to examine the problem of singularities in ML estimation of a mixture of two normal regressions. As expected; results indicated that the incidence of singularity decreases with the increase in sample size as well as with the increase in the angle of separation. A simple method of generating starting values was also proposed.

引用

页码：667 / 674

页数：8

共 12 条

[1]

[Anonymous], 1974, COMMUN STAT-THEOR M, DOI DOI 10.1080/03610927408827201

[2]

BEARD TR, 1991, REV ECON STAT, V75, P654

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