I would like to create a composite scale for different attitudes by calculating the mean score from different items on a questionnaire. Most of the items use Likert scales however some of the items on the questionnaire just use yes/no answers. Is it still possible to combine these scores to create a composite scale? How would I do this?

e.g. Question 1 Likert scale 1-5 Question 9 Yes/No

  • $\begingroup$ It's certainly possible; the questions are more 'how' and 'does it have the properties you want'? Likely you'll want some linear combination, and the obvious thing is to have a set of positive weights (or instead, weights corresponding to the 'direction' the question works in, if they're not all going the same way). The question comes down to how to give the last question 'enough' weight relative to the Likert items, but not have it unduly dominating. The other thing to note is you probably won't end up with a good approximation of a normal distribution for your score (may not be a problem) $\endgroup$ – Glen_b May 28 '13 at 11:33
  • $\begingroup$ So if I coded Yes as 5 and No as 1 relative to the Likert scale would this be acceptable? I am conscious of skewing the data too much either way by assigning too high values for each item. $\endgroup$ – Samantha Caffull May 28 '13 at 14:16
  • $\begingroup$ Well, it seems fairly reasonable, but it's not me you need to convince. $\endgroup$ – Glen_b May 28 '13 at 15:12

You can do this, but taking the simple mean is not good anymore because of different scales. You can do factor analysis to get at the "latent construct" you are trying to evaluate, or principal components for pure data reduction. There's been work on doing these with a mix of variable types.

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.