Have an idea?

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

Number of versions for Sparse Max Diff

When using LightHouse Studio to generate a sparse max diff design where you allow individual designs lacking connectivity - are 100 max diff versions enough to be able to run HB analysis to obtain individual level max diff scores (or do you need 200 or the default 300)? 100 versions would seem sufficient but does Sawtooth have any tests/best practice recommendations on MINIMUM number of versions needed for HB analysis with Sparse max diff?
asked Jan 8 by anonymous

1 Answer

+1 vote
This question really depends on how many items there are, how many items are shown per set, how many sets shown per respondent.  So, it's hard to generalize.

In my opinion, with a sparse MaxDiff design where each respondent saw each item 1x, as few as six versions should be sufficient to approximate the value of having 300 versions under HB estimation.   

But, to be certain, one would need to run a parameter recovery experiment with robotic respondents under different design conditions.
answered Jan 8 by Bryan Orme Platinum Sawtooth Software, Inc. (172,790 points)