1

I would like to reduce the amount of data I have before conducting an analysis.

My data consists of several sets of questions assessing comprehension of various topics (e.g. 4 questions assess comprehension of one concept, 4 questions assess comprehension of another concept and so on). I would like to sum the questions for each concept before running further analyses. But, I am not sure whether it is ok to do this without first conducting a factor analysis or some other reduction technique?

If I do go down the road of using a multivariate method like FA to reduce my data, then I run into another problem. My data is categorical and so I don't know whether ordinary FA is suitable to use. I see there is the option of Categorical Principal Components Analysis. But, I am not sure if this is suitable. How do you think I should proceed?

Salada
  • 11

1 Answers1

4

You can add a simple sum of your items to your model, but you would need to justify it. Adding the sum would be equivalent to adding all your items but constraining the effect of each of these items to be the same. This might make substantive sense, or it might not. It is certainly a testable constraint. See, for example: M.L. Buis (2012) "Stata tip 108: On adding and constraining", The Stata Journal, 12(2), pp. 342-344. http://www.maartenbuis.nl/publications/sum_constr.html

If you are looking for a more formal model, then, given your brief discussion of your items, it sounds as though an IRT (Item Response Theory) model would be more appropriate than a factor analysis.

Nick Cox
  • 56,404
  • 8
  • 127
  • 185
Maarten Buis
  • 21,005