I generated a Likert scale survey to test perceptions of quality, authority and usability on a website. No problems there. Each of the items above is constructed from 5 dependable variables, so I intend to use MANOVA for each one, again no problem. Right..

Each of the 5 dependable variables is given a score per person between 4-20, generated from the sum of four questions (i.e. each question gives 1-5 to the total score).

So my Likert scale is nothing too special. Here is where I am stuck. :shakehead

Because of problems getting enough participants in piloting the survey I was not able to do factor analysis at the development stage. So I have read I need to do confirmatory factor analysis. But I am stuck on how to cut up the date. Do I:

1 - Do confirmatory factor analysis on the dependable variables (i.e. the 15 that exist in the whole survey to ensure that they do fall into the three categories I predict, or to produce 2, 3 or 4 categories).

2 - Do item factor analysis on all 60 questions of the survey, to pull out the dependable variables from those, and then carry on with the 3 categories for MANOVA

3 - Do item factor analysis followed by confirmatory factor analysis (then MANOVA on the outcomes)

4 - something completely different

NB: In case it helps, I am using SPSS 19