HI: I know there are couple of questions relating to comparing the same survey item with differing numbers of response categories, but I did not see a question that dealt specifically with this issue.
Namely, I want to compare the dispersion of two sets of survey responses on 12 items per survey. The items on Survey 1 use a 6 point Strongly agree- Strongly Disagree scale, while the items on Survey 2 use a 4 point Strongly agree to Strongly disagree scale.
My gut tells me that I cannot use standard deviations to make this comparison to assess dispersion from the mean. Just by virtue of the fact that Survey 1 provides more categories, it should have more dispersion. Could I convert both items to z-scores to compare the dispersion? Or should I try to rescale the responses? If so, should I rescale from 1-6 to 1-4 or go the other way?
This is not for peer-reviewed research at the moment; I have the working hypothesis that the context in which survey 1 was administered is marked by greater disagreement on the items that are measured while there is less conflict and more agreement in the context in which survey 2 was administered.
I guess another option would be to compare the percentage of respondents who picked the most extreme options.
Thank you for your suggestions.