How important are zero-centered utilities compared to raw utilities?

Question

Hi all,
I have been using Sawtoth for a research project and want to publish the results in a scientific paper. For my research purposes, the preferences shares are the most important results. However, I would also like to include the utility results and explain the findings drawn from them. I used HB for the analysis.
Are zero-centered or raw utilities best for this?
Since the preference share model is based on the raw utilities and they are also zero-centered, I think it would make sense to show them. However, in most of the literature I find the "zero-centered utilities". What are the advantages of zero-centered utilities?

Thank you very much!

Answer 1

Check with your reviewer.  They may prefer to see the utilities reported on a raw scale rather than on the larger (normalized) scale as our software does.

Normalized (zero-centered diffs) scale puts each respondent on about the same footing toward impact on the mean.  But, the raw scale is the original logit scale as coming out of HB-MNL estimation.