Multivariate extensions of univariate measures of association that are employed in comparative experiments are reviewed. Two of these are generalizations of the correlation ratio, and the third is a generalization of Hays' omega. The numerical values of these three indices are empirically compared along with a fourth index, the proportion of correct classifications, using two sets of data and various numbers of predictor variables. As in the univariate case, the values of the correlational indices are nearly the same in all situations investigated. The use of a correlational index versus the proportion of correct classifications depends upon the intent of the researcher.
Download full-text PDF |
Source |
---|---|
http://dx.doi.org/10.1207/s15327906mbr0704_8 | DOI Listing |
Enter search terms and have AI summaries delivered each week - change queries or unsubscribe any time!