I have 130 or so rows of questionnaire data. It is 67 columns wide mostly of categorical nature (and often yes/no). I have been using daisy, pam and plot to view cluster plots with subsets of columns. With different subsets, there is great variation in the degree of overlap of the clusterplot ellipses, which I assume indicates some measure of differentiation between clusters. Can anyone point me at forms of analysis which can help identify column subsets which yield "high" clustering (separation of ellipses?)? Sorry for the verbose description, there is probably a single word or phrase which describes what I would like to do. I just don't know it. Ultimately I am trying to identify distinct groups within the respondents and find what areas (columns) that define the groups. cheers