Have an idea?

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

From HB utility estimates to simulated shares of preference

This feels like a silly question...

If I use the utility estimates from a Logit analysis, I can calculate the shares of preference by hand and get the same values as the simulation software. P(A) = exp U(A) / [exp U(A) + exp U(B)].

If I use the utility estimates from HB, my calculations by hand don't match those provided by the simulator in the software.

The orders of magnitude are different when using the HB estimates. So, do they need to be transformed before plugging them into: P(A) = exp U(A) / [exp U(A) + exp U(B)]? Or what is the "back end" formula to get from the HB estimates provided by the software to the simulated shares of preference provided by the software?

asked Nov 22 by anonymous

1 Answer

0 votes
Our simulators default to a method called Randomized First Choice, which runs a simulation a ton of times and adds a bit of special error to the respondent utilities in each iteration.  If you change the simulator to use Share of Preference as the simulation method, you should be able to get the same numbers as doing it by hand (make sure you are running the simulations per respondent and not on the average utilities).
answered Nov 23 by Brian McEwan Platinum Sawtooth Software, Inc. (54,165 points)