ABSTRACT
The k-means procedure is an attempt to minimize the variance within each cluster and consequently to maximize the variance between clusters. It is one of the few programs that allows for testing hypotheses. It allows the researcher to prespecify the number of groups of objects that will be delineated. K-means clustering is part of the SAS packages of programs under PROC VARCLUS. Donovan attempted to classify children into three learning modality groups. These were Auditory Preference (AP), Visual Preference (VP), and No Preference (NP). She used a battery of ten diagnostic subtests from the Illinois Test of Psycholinguistic Abilities and the Gates MacGinitie Readiness Skills Test. Donovan specified a clustering into three groups without specifying any basis for the clustering. A k-means iterative clustering using a computer program developed by McRae called MICKA used the minimization of the within cluster variance and distance criterion was Mahalanobis' d2.
