Have an idea?

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

Counts analysis 2-way interactions vs. Interaction search

Dear all,

I have a question regarding the 2-way interaction effect detection included in the counts analysis and the interaction search method: Could you tell me what exactly the difference is between those two approaches?
I am asking, because when looking at the 2-way interaction effects in the counts analysis, none is significant; however, when searching for interactions by using the interaction search, two different combinations of attributes sho significance and even improve my model.

Thanks for help :-)

Best regards
asked Jul 27, 2017 by briniminii (490 points)

1 Answer

0 votes
Best answer
The counts analysis doe not take any utility model into account - it just looks at the distribution of responses and does a chi-squared test to see if the distribution is significantly uneven or not. It is intended as a quick-and-dirty way of checking for interactions.  

The interaction search is the more accurate way to go because it does the appropriate formal statistical tests, but be careful to account for experimentwise error when using it:  https://www.linkedin.com/groups/1715557/1715557-5895943758326820864
answered Jul 27, 2017 by Keith Chrzan Platinum Sawtooth Software, Inc. (93,025 points)
selected Jul 27, 2017 by briniminii