I would greatly apreciate some advice on selecting an appropriate form of factor analysis. Im writing a paper that introduces a new questionnaire. This new questionnaire is an age-adapted and translated version of an older and well used/established questionnaire.
The point of the paper is both to validate the questionnaire for use on this new age group, and also to utilise a different method of calculating the results. The original questionnaire measures two different sub-scales ( two theoretical dimensions). On theoretical grounds there is good reason to try to measure according to three dimensions (three different subscales). Indeed i already have a theoretically derived method of calculating three sub scales. This method of calculating the questionnaire have already showed to give results in theoretically expected ways, which in part validates this method.

Beyond this id like to perform some form of factor analysis that gives more empirical support to this theoretically grounded new way of calculating the scores.

I guess my question mostly is this: Can i simply perform an exploratory factor analysis, forcing three factors, using the pattern matrix as a new coding scheme and then interpret the eigen values/ scree slope and report loadings as support for my model?

Or would it be far better to use a confirmatory factor analysis? If so whats the best way to go about performing one?

Im using SPSS software, and would prefer to use it.

The data is N≈4500. Items are Likert scales, some of them definately to skewed to be called normally distributed.

Any input regarding this, what methods to choose and so on is apreciated!