Have an idea?

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

Subset of MaxDiff Scores In Turf

Is it statistically valid to to use a subset of your maxdiff items scores in the subsequent TURF analysis?  For context, we have 45 new program features evaluated in the MaxDiff.  Of the 45, 15 are very easy to implement.  We want to see the best combinations out of the 15 easy to implement features.  I am easily able to prohibit the 30 items that I do not want to be included in the TURF using  ' Items that Cannot Occur Together in Portfolio' window of the TURF and limit the potential items to the 15 I want in my TURF analysis portfolio.  My question is if this is a valid method to use and if the results are meaningful since some items. It seems okay, since all items were compared to the same items in the maxdiff in an on-biased way, but I would like confirmation.
asked Feb 24, 2016 by Joe

1 Answer

0 votes
MaxDiff assumes the items may be dealt with independently.  That's the nature of the experimental design (each item appears an equal number of times with every other item so that the independent, unbiased effect may be measured), the utility estimation, and that notion can naturally be carried forward to TURF.

In reality (and this is more a realm for academic pursuit), the utility of an item may be context-dependent and thus dependent on the items it is being compared to.  But, such context effects within MaxDiff are typically very tiny (relative to the independent score for each item) and virtually always ignored by practitioners.
answered Feb 24, 2016 by Bryan Orme Platinum Sawtooth Software, Inc. (132,290 points)