Have an idea?

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

Express Maxdiff study attributes number

I am reading the below topic from CBC/HB Help and confused in (K-1) x (k-1) specifications while testing.

Appendix K: Estimation for MaxDiff Experiments: "To avoid potential problems when dealing with dummy coding under HB and very sparse data conditions, you should specify a more appropriate "custom prior covariance matrix" from the Advanced Estimation Settings tab.  Check the custom prior covariance matrix box and click Edit....  Specify a (k-1) x (k-1) matrix of values, where k is equal to the total items in your MaxDiff study, a k-1 is equal to the total number of parameters to be estimated (also equal to the number of attributes in the Attribute Information tab).  If you wish to specify a prior variance of 1.0 (a typical default), the matrix is composed of "2"s across the main diagonal, and "1"s in the off-diagonal positions, such as:"

I have 20 MaxDiff Exercise showing 30 attributes in each exercise. Respondent will see only one exercise and then move out of MaxDiff questions.

I have total 100 items and in the attributes tab 99 (k-1) listed. If I understand correctly the second (k-1) refers 100 attributes right?

Thanks in advance,
asked Sep 24, 2013 by Lawrence Bronze (2,100 points)

1 Answer

0 votes
If you've got everything coded properly, the prior covariance matrix should be 99x99.
answered Sep 24, 2013 by Bryan Orme Platinum Sawtooth Software, Inc. (181,965 points)
Hi Bryan,

I did the same and looking for the results.

Follow-up question - Maxdiff study attributes