Have an idea?

Visit Sawtooth Software Feedback to share your ideas on how we can improve our products.

Comparing differences in Maxdiff utilities

I understand that in a Conjoint analysis, I can compare the differences in utilities for two attributes i.e. if we have shape and size and the difference between the utilities of blue & red is 20 and that between square & round is also 20, we can say that for that particular respondent, blue is just as preferred over red as square is preferred over round.

However, I was wondering what would be the watch-outs of comparing the respondent-level utilities if Maxdiff was done separately for the two attributes one after the other by the same set of respondents i.e. the respondent remains the same but the estimation happens separately for the two attributes?
asked Jul 6, 2014 by Aritra Kanjilal

1 Answer

0 votes

I think there are watch-outs, red flags and bright flashing warning signs on this one.  If I understand correctly you want to compare items from two separate MaxDiff exercises (please correct me if I do not understand you correctly, but it sounds to me like this is what you're doing).  If so this is no better than comparing part-worth utilities from two separate conjoint experiments.  In either case, even if the respondents are the same, the context effects of whatever other attributes/levels/items are shown (and I'm assuming these are different in the two experiments) prevent you from directly comparing part worth (or item) utilities estimated in two different models.
answered Jul 7, 2014 by Keith Chrzan Platinum Sawtooth Software, Inc. (93,025 points)