THE USE OF MULTIVARIATE REGRESSION USING DUMMY VARIABLES IN SOCIAL-PSYCHOLOGICAL EXPERIMENTS.
Abstract
Multiple regression analysis using dummy variables makes regression techniques possible in experimental designs instead of analysis of variance. These techniques have been applied previously to demographic data, but not to psychological experiments. Dummy variables are created by 'scoring' individuals according to the presence or absence of a dichotomous trait, e.g., '1' or '0' for being in an experimental group or not. Several categories are dealt with by creating n-1 dummy variables leaving one category with a zero value on all dummy variables. In a multidimensional design, dummy variables are created for each dimension, for each combination of categories or particular combinations of dimensions. No assumption is necessary in regression analysis about variables underlying the qualitative categories. Dummy variables can be combined in the regression equation with continuous variables or other dummy variables representing background or previous experimental treatment. In this case the slope of the regressions of the dependent variables on the continuous variable can not be significantly different in the separate sectors of the population defined by the dummy variables, that is, no analysis of covariance problem should exist. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1966
- Accession Number
- AD0640568
Entities
People
- H. H. Winsborough
- Kurt W. Back
Organizations
- Duke University