Based upon previous work by Fouts (1987, 1989, 1992), the statistical procedure
chosen for this study was cluster analysis (CA). CA employs the opposite process of
discriminate analysis,
in which the researcher begins with well-defined groups
and
attempts to discover the variables that account for differences among groups. In cluster
analysis, the researcher begins with one large, undifferentiated group and forms
subgroups that differ for selected variables (Kachigan, 1991).