A SET OF PROGRAMS FOR COMPUTING MULTIPLE CORRELATIONS AND RELATED STATISTICS: TWO ANALYSIS-OF-VARIANCE PROGRAMS FOR DISPROPORTIONATE CELL FREQUENCIES,
Abstract
A stepwise linear regression and multiple correlation program, with distinct accretion and deletion phases, has been developed for use with either complete or incomplete data. A companion cross-validation program makes it possible to test the derived regression coefficients by applying them to a different sample obtained from the same parent population. The difference between the multiple correlation and the cross-validity indicates the stability of the regression coefficients. The set of programs offers great versatility and high capacity. In theory it has been known for many years that by using the general linear model, multiple regression can give results that are identical to those of analysis-of-variance, (anova). In particular, multiple regression can give an anova for disproportionate cell frequencies. Program GENIV does this for one-way, two-way, or three-way anova. Program DEVIANCE performs two-way anovas, automatically selecting every possible pair from a given set of predictors. Both programs generate all possible interaction variables. They binarize, (dichotomize), graded or ungraded raw predictor scores, assigning each subject to the specified levels of each factor. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1968
- Accession Number
- AD0670592
Entities
People
- Ardie Lubin
- Cyril H. Nute
- Donald D. Beck