09-04-2014 04:07 PM
I want to use PROC CORRESP (MCA option) as a pre-processing for clustering. The purpose of this is to transform categorical variables to continuous variables. Does anyone have any sample codes or examples for this? How can I create output dataset from CORRESP (MCA option) and then do the hierarchical clustering with ward criteria?
09-05-2014 10:22 AM
I don't think you can use PROC CORRESP to transform categorical variables to continuous variables. It decompose cell chi-square into X axis and Y axis , therefore you can make some cluster idea,but that is only for CATEGORY variables.
09-05-2014 01:14 PM
I can kind of see a clunky way of doing this. After PROC CORRESP does its thing and you get the X and Y axis values, you could use PROC DISTANCE to get a single continuous value, or just use the X and Y axis values as continuous variables.