ABSTRACT
Contingency/frequency tables are the most straightforward type of data to handle by dual scaling discussed in this book. Let us first recapitulate what dual scaling will do with this data type. In summary form, dual scaling will determine weights for the rows, y i , and weights for the columns, x j , of the contingency or frequency table in such a way that:
The row-column association in the data will be maximally explained;
Analysis will continue, as necessary, to extract more dimensions to explain other distinct patterns of association;
The correlation between the rows and the columns will be a maximum;
The regression of y i , on x j , and the regression of x j , on y i , will be simultaneously linear; and
Both between-row and between-column discriminations (variations) will be maximal.
