Have an idea?

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

Using MaxDiff in other analysis


Is there a freely available Sawtooth technical paper or webinar covering best practices in how to use max diff scores (that is, the 100 point scores typically reported to clients) for subsequent multivariate and other analysis - such as for correlation, factor analysis, regression, discriminant analysis, etc. that are often done utilizing 5 or 10 point scale survey items?

Thank you.
asked May 10, 2021 by anonymous

1 Answer

0 votes
I am not aware of such a white paper, no.  

There are examples in the academic literature of folks using MaxDiff scores in subsequent analyses (I presented a paper on this topic at our conference last month so send me an email at keith@sawtoothsoftware.com and I'll send you a list of references, as well as a paper we published in Quirks magazine on the topic).  I know I've used raw MaxDiff utilities (not the sum-to-100 ratio scaled scores you're referring to) in correlations, factor analysis, reliability analysis, cluster analysis and perceptual mapping, and the academic papers would add multidimensional scaling to that list.
answered May 11, 2021 by Keith Chrzan Platinum Sawtooth Software, Inc. (111,275 points)
Thank you for the very helpful response. Would it be possible to please explain why you use and recommend using raw MaxDiff utilities (not the sum-to-100 ratio scaled max diff scores) for subsequent multivariate and other analysis?
The raw utilities (or the zero-centered diffs scores that are partially corrected for differences in response error) are the MaxDiff measures that align best with the original best-minus worst scores that have mostly been used for subsequent analyses in academia.  I do not yet know how well the probability transformed scores (which lend themselves to easy interpretation and reporting of the items' relative positions) will perform as inputs to multivariate analyses.