ABSTRACT
The idea of a classification analysis based on multidimensional data is to group the individuals according to their value profiles. Usually it is then assumed that the variables are approximately continuous and can be treated as being on an approximate interval or ratio scale. We saw in the preceding chapter that cluster analysis could then be used to accomplish the classification. However, one or both of the following two conditions can apply in a specific context:
The scale levels of the variables do not permit the use of a standard cluster analytic approach based on interval scale data (e.g., some variables are on an ordinal scale or are pure categorical variables).
There are reasons to believe that the clusters resulting from a standard analysis will not give an adequate representation of the individual profiles.
