I was curious if Factor Analysis of a Likert Scale questionnaire in an effort to pull out common sets of questions is possible. My statistics teacher thinks this is not valid because the data is discrete. However, I believe that you can still find top few eigenvectors which maximize the variance and look at the loadings in each vector to see which questions comprise the linear combination. The questions involved can then be grouped to be a "common set". Is this type of exploratory analysis valid? Any thoughts about using clustering analysis instead? Thanks in advance for your input.
Advertise on Talk Stats