Have an idea?

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

Do "zero-ceneter diffs" utilities mean you can graph all utilities on one graph?

I'm using the Analysis Manager within Lighthouse studio and used the average utilities (zero-centered diffs) after doing an HB analysis. Does this mean that all the utilities are now normalized and I can graph utilities for all levels and attributes all on one graph? In other words, am I able to compare the utility of Level 1 Attribute A with the utility of Level 4 Attribute C?

Also, this is a more general question: When would it be more appropriate to use Logit, as opposed to HB analysis?

Thank you!
related to an answer for: Analysis on Utilities
asked Apr 26, 2017 by anonymous

1 Answer

+1 vote
No.  Transforming utilities via ZCD makes them more comparable from one respondent to the next (because it attempts to correct for differences in consistency among respondents).  But each attribute is still zero-centered, and the zeros are arbitrary, so you still cannot make direct cross-attribute level comparisons.

To you second question, we often use logit as more of an exploratory analysis to investigate model structure while we use HB for the final analysis we'll use in reporting and simulations.  Logit would be as appropriate as HB appropriate if (a) the IIA property of proportional share draws applied to your data and (b) you were not interested in doing subgroup analyses; otherwise HB would probably be the better bet.
answered Apr 26, 2017 by Keith Chrzan Platinum Sawtooth Software, Inc. (93,025 points)