Have an idea?

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

Express Maxdiff with None of these?

Hi

I wonder if we could implement "None" option in some MaxDiff exercises the same way we use (and later interpret) "None" in CBC. Specifically, such "None" could be useful for Express MaxDiffs, with high (40+) number of alternatives, where often we ask only about best choice anyway and for which an implementation of Anchored MaxDIff would render the module way too lenghty for a consumer to answer with full focus.

I have an idea how to manually implement it, but before diving into tests I wanted to check if anyone tried running MaxDiff on Best-only with None? If yes, then what was the approach assumed? If no then what kind of technical issues could that imply for Sawooth Software?

Thank you very much in advance
asked Mar 16, 2018 by Piotr

1 Answer

0 votes
Hello, Piotr,

I have tried MaxDiff with None.  It can require some careful wording of the None alternative.  I've tried it both with an always-present none alternative as well as one that pops in and out of choice sets as one of the attributes.  

I was never able to do with with both bests and worsts because of the difficulty of wording the None alternative to cover both cases.  

I'll say that I haven't done this at all since anchored MaxDiff came out, because that has always worked better for my needs.
answered Mar 20, 2018 by Keith Chrzan Platinum Sawtooth Software, Inc. (71,325 points)
Hi Keith

Thanks for your reply. Could you cast some light on approach you assumed when running MaxDiff with None always present?

Specifically, in light of your experience, would that work to create a rotation file for n items, add "None" manually to the rotation file prior to field start and later run all the analyses as if n+1 were programmed from the very beginning?

Regards
Yes, I believe that is how I did it.
...