Have an idea?

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

Calculation and interpretation of interaction effects


we have a design with 5 attributes. One attribute is categorical with two levels (brand 1, brand 2) and 4 attributes are ordinal with 3 levels each (e.g.  degrees of severity). We want to estimate interaction effects between each brand and the levels of all other attributes.

Now we are trying to replicate the results obtained from cbc-logit analysis (using SMRT) in Stata but failed. How are interaction effects calculated in Sawtooth?

As Sawtooth uses effects coding how do I have to take this into account when interpreting interaction effects? E.g. when brand 1 (level 1) interacts with level 1 of another attribute and brand 2 (level 2) with the same attribute-level, is the coefficient of the interaction term of the latter the inverted sum of the coefficient for the brand 1 interaction term?

Any help is appreciated.

asked May 12, 2015 by Andrew Bronze (975 points)
retagged May 12, 2015 by Walter Williams

1 Answer

0 votes
Please write me at my email at "bryan at sawtooth software dot com" and I will share some PowerPoint slides with you on this.
answered May 12, 2015 by Bryan Orme Platinum Sawtooth Software, Inc. (148,140 points)

I sent you an email but not sure whether it reached you.

Thanks for helping out.

I'm sorry, I never received your email.  Please write me at bryan at sawtooth software dot com and I'll get you that information.