A comparison of multiple regression computer programs and their usefulness in analysis of variance

UNCG Author/Contributor (non-UNCG co-authors, if there are any, appear on document)
Douglas Lane Mills (Creator)
The University of North Carolina at Greensboro (UNCG )
Web Site: http://library.uncg.edu/
William Powers

Abstract: This research will examine the multiple regression programs from major statistical packages and document for each the statistical procedure used, the cost, the difficulty of use, and the output. Each program will be utilized for analyses of variance in balanced and unbalanced designs with an interest in determining the accuracy of and the statistical techniques employed by each package. The regression procedures, while using dummy variables, are superior to the ANOVA subroutines for the solution to ANOVA. All four packages studied have regression programs that can be used in an unbalanced ANOVA. The REGR procedure in SAS is best for the standard form of regression, while SPSS and BMD are best for stepwise regression. All packages have the limitation of not giving all types of correct analyses. Therefore, Appendix 4 gives an ANOVA program written for the Programming Language One Optimizing Compiler for a two factor design that will give five correct analyses.

Additional Information

Language: English
Date: 1977
Regression analysis $x Computer programs

Email this document to