Have an idea?

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

Distribution of Individual-Level Utilities from HB

Thank you very much. This was really helpful.

Can I refer to any Research oder technical paper for the last part regarding the interaction effects? I thought that with main effects one cannot Interpret the main effects of the attributes involved. However, the interaction effects can still be seen as a correction of the main effects, right?
asked May 28, 2016 by anonymous

1 Answer

0 votes
I'm sorry, but I cannot remember right now any literature reference for my statement regarding effects-coding in conjoint designs leading to the ability to interpret the main effects independent of the interaction effects.  I think I once read it in a Joel Huber paper, but after some search of our website and reviewing some older Sawtooth Software conference proceedings, I still cannot find it.
answered May 29, 2016 by Bryan Orme Platinum Sawtooth Software, Inc. (172,790 points)
Thank you very much for searching the paper. If you find it, please tell me since it would really help me. I am still not sure how to present the average ZCD from HB estimation. I found one interaction between brand and price. Now I am not sure if I can interpret the values from the main effects of brand and price without taking into account the interaction effects. Moreover, how should I interpret the values from the interactions?
If your attributes are independent of one another in preference (utility), then main effects alone will reflect the relative preferences for levels within the attribute and the relatively impact (importance) of each attribute on product choice.  But, if there is an interaction between brand and price, then the utilities for price (the importance for price, or the shape of the price curve, or both) depend on which brand the prices are attached to.  That means that to assess the impact of price on choice, it would be more accurate if you knew which brand you are referring to.

But, if you want to capture as a summary measure the average impact and preference of brands holding all other attributes constant (at the average of all the other levels of other attributes) or the average impact of price (holding all other attributes constant including considering the average effect across all brands) then the main effects from a main effects + interaction effects specified model will tell you what those summary (main effect) measures are.  It will, indeed, be something of an abstraction, since it doesn’t refer to a specific brand or a specific price, but an average across them.

Then, the interaction effects that remain after accounting for the main effects are the specific adjustments to total preference that fit the data when a given brand is coupled with a given price.