Have an idea?

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

Does it make sense to first use LC to run a segmentation, and then use the classification as covariate in the HB run?

This would apply for both MaxDiff and conjoint exercises.
asked Nov 19, 2014 by anonymous

1 Answer

0 votes
This question has been raised many times over the last few years.  

Formal Bayesian statisticians do not like this procedure, as it is said to be "double-dipping" into the data wrt heterogeneity.

I have heard some practitioners say that although they know it isn't formally proper to do this, they have done it and see good results.

So, proceed cautiously and recognize that if you do this you are doing something that isn't formally proper.
answered Nov 19, 2014 by Bryan Orme Platinum Sawtooth Software, Inc. (139,965 points)
...