Have an idea?

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

Best-Worst CBC estimation algorithm

I'm analyzing a best-worst CBC with CBC/HB software. In my master thesis I need to write down how the software estimates the part-worths. I found information regarding the bayesian analysis, Gibbs-sampling and Metropolis Hastings Algorithm in the CBC/HB manual.

Does the software use these methods as well for best-worst?
asked Nov 8, 2015 by Sabrina (380 points)
retagged Nov 8, 2015 by Walter Williams

1 Answer

0 votes
Yes, Sabrina, it is.  The analysis is the same for best-worst CBC only the design matrix is modified.
answered Nov 8, 2015 by Keith Chrzan Platinum Sawtooth Software, Inc. (90,575 points)
What do you mean with design matrix? Is it the matrix of variances and covariances of the multivariate normal distribution?
No, the design matrix is the matrix that describes the concepts and the choice sets.  It has as many rows as there are concepts x choice sets x respondents.  Each row has a description of the concept (columns are effects-coded) plus a response variable indicating whether it was chosen or not.  

In Best-Worst we double the tallness of that matrix by stacking beneath it a negatively coded (all columns for effects-coded concept descriptions multiplied by -1) matrix plus response indicators for the items selected as worst.
Thanks for the answer!