Have an idea?

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

Comparing Willingness-to-pay


I've been able to calculate the WTP based on the average zero-centered utility values for three different groups with a Hierarchical Bayes Estimation. Now I want to know if they are significanty different from each other. How do I compared them in a formal testing?

Best regards

PS: I already read the dangers of calculating WTP in the sawtooth paper.
asked Jul 4, 2017 by anonymous

1 Answer

0 votes

Your respondent level WTPs are metric variables like any other.  However, because they are based on calculations involving division, they are more subject to outliers than would be a rating scale.  As a result, I would probably NOT just subject them to a simple t-test like you learned in school.  Instead I recommend one of two things:

You could conduct a median test using the point estimates - this is a non-parametric version of a t-test that's more appropriate to run when your distribution could include outliers.  

You could also run a more Bayesian version of a t-test.  This involves getting into the draws file but the test itself involves no more than counting.  Please get in touch with me at keith@sawtoothsoftware.com if you want more information about how to do this test.
answered Jul 5, 2017 by Keith Chrzan Platinum Sawtooth Software, Inc. (50,675 points)