Have an idea?

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

Test for significant difference in linear estimates of two different attributes (HB Model)

Dear Sawtooth Team,

I conducted a CBC analysis with various attributes. Two of these attributes are "numeric" with 6 levels each and equal distances.

I used the hierarchical Bayes model (HB) from Sawtooth software to estimate the part-worth utilities and coded the quasi numeric attributes as linear term (taking into account that I specified the magnitude of about single digits prior to running the model).

Now I got two plausible linear parameters.
I was wondering, if there is a way to test for significant difference between the two linear coefficients. I.E. that the one parameter leads to a signficant higher decrease (increase) in utility than the other parameter by increasing (decreasing) one level?

Is it appropriate to conduct a paired t-test?

Thanks in advance.
Best regards,
asked Jan 28 by meschusti (125 points)

1 Answer

+1 vote
Yes, doing a simple paired t-test is how I think most folks would test that hypothesis.
answered Jan 28 by Keith Chrzan Platinum Sawtooth Software, Inc. (102,700 points)
Great, thank you.
Another way to do this is to rely on Bayesian inference, as Bayesian statistic does not rely on asymptotic properties as classical statistic does. Therefore, one way to go is to use  probability intervals of posterior samples. As Lighthouse studio allows you to save the random draws of the estimation phase, this is an easy thing to do.