Questionnaire analysis


New Member
I feel a bit strange posting here, but everything's worth a try! I am trying to analyse a questionnaire that I designed myself (sadly, my design was a bit thoughtless, as I have been dumped in a psychology department without any prior knnowledge of anything to do with psychology - least of all their research methods!) ...

I have an IV (A - 20 levels), 4 co-variants (a - 13 levels, b - 13 levels, c - 2 levels and d - 3 levels) and two types of response: (1) a Likert scale score (5 point), and (2) a YES/NO (1/0) response. I want to know is there are differences in the responses between the levels of A and what effect the co-variants have on this difference.

I'm thinking some kind of logistic regression would be appropriate, because I've read that they can deal with the binary response format and account for co-variants. Does that sound feasible?

I've never done anything other than basic non-para stats!

And what do I do with the Likert scale? Can that somehow be included or would I in any case have to analyse the two seperately?

I'm also wondering about running a principle components analysis on response (1) in order to group the responses (that would give me factors, right? Would I then be able to analyse those factors instead of testing the binary variables directly?) ... but then what? I'm full of ideas, but sadly without much knowledge.

Any thoughts would be appreciated, I'd really like to get my head around this!