I am working on developing a survey for the purpose of cluster analysis and needed some help conceptually or methodologically, The survey has, let’s say, 200 questions. I know that many of these are redundant and we can most likely cut out 195 of the questions and perform the cluster analysis on only 5 questions and get the same (or close to the same) clusters. I was wondering what methodology I would use to cut down these questions and determine which, let’s say 5, are the most important in developing these clusters.

It is important that our clustering scheme has as few questions as possible but at the same time has the same vigor as a scheme developed from 200 questions.

I recognize this is most likely related to factor or principle component analysis but I do not know how these would work into developing a solution.

Thank you very much