Have an idea?

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

HB estimation of BPTO

Has anyone tried estimation of classic BPTO data with CBC/HB? If so, could you share any suggestions?

regards
asked Oct 27, 2011 by anonymous
retagged Sep 13, 2012 by Walter Williams

1 Answer

+1 vote
I've never tried this, but one way that occurs to me is to code each response as a choice task.  In each "task" a concept (brand + price) was compared with other concepts (brands + prices) that were "shown", and the selected concept was "chosen".  You could code the brands and prices both as main effects.  And, you could also look into estimating interaction effects between brand and price (though they probably will not be very significant).

On the other hand, HB estimation really isn't needed to develop a simulator.  The BPTO provides a full rank-order of all brand x price combinations.  So, you can develop a "first choice" simulator simply by looking at the concepts in the simulated scenario, and indicating (for each respondent) which concept has the highest observed ranking in the data.  That concept is the "first choice."

But, if you terminate the BPTO early, before the respondent has fully ranked all cards, then the final task would show that the respondent picked the last card selected within a set containing ALL remaining cards (not just the cards that are being "shown" in the set).  In this case, the "not selected" brands and prices will receive lower net utility than the last selected brand + price.
answered Oct 28, 2011 by Bryan Orme Platinum Sawtooth Software, Inc. (131,390 points)
edited Oct 28, 2011 by Bryan Orme
...