RECRES - A FORTRAN-77 PROGRAM FOR DETECTING MODEL MISSPECIFICATION IN MULTIPLE LINEAR-REGRESSION

被引:0
|
作者
BATES, BC
SUMNER, NR
机构
关键词
MODEL MISSPECIFICATION; MULTIPLE LINEAR REGRESSION; RESIDUAL PLOT; RECURSIVE RESIDUALS; QUANDT LOG-LIKELIHOOD STATISTIC; NORMAL PROBABILITY PLOT; CUMULATIVE SUM PLOT;
D O I
10.1016/0098-3004(91)90081-N
中图分类号
TP39 [计算机的应用];
学科分类号
081203 ; 0835 ;
摘要
RECRES is an interactive FORTRAN-77 program for detecting model misspecification in multiple linear regression problems. The program produces plots of ordinary least-squares (OLS) residuals, normal probability plots of recursive residuals, cumulative sums of recursive residuals, the Quandt log-likelihood statistic, and recursive estimates of the model parameters. It uses orthogonal triangularization to compute diagnostic statistics and allows flexibility in the production of diagnostic plots. Examples show that the program yields reliable results and can detect model misspecifications that are not apparent readily to users of conventional multiple linear regression programs. A source listing of the program and an example of an interactive session are provided.
引用
收藏
页码:91 / 114
页数:24
相关论文
共 50 条