Have an idea?

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

How should I use interaction effects when calculating share of preference?

I had a very general question.  There may be a quick answer, or direction to some other technical papers you have. Essentially, I have seen (in calculating share of preference for traditional CBC exercises) main effect utilities be summed WITH interaction effect utilities.  I have also see interaction effects used on their own (main effects ignored).   Is there a theoretical basis/justification for calculating share of preference one of these ways vs the other?
asked Apr 14, 2017 by JKincaid Bronze (800 points)
retagged Jun 19, 2017 by Walter Williams

1 Answer

+1 vote
I have rarely seen cases where we estimate and use interaction effects without the main effects.  The cases where I have seen it were unusual ones, so I don't want to say it's impossible.  The theoretical justification usually supports using main effects, because the interactions by themselves usually aren't that meaningful except in the context of the main effects.  

Please feel free to provide more details about study specifics and perhaps I can be more expansive in addressing your question.
answered Apr 14, 2017 by Keith Chrzan Platinum Sawtooth Software, Inc. (50,900 points)