Have an idea?

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

Utility scores and constructed lists ACBC


For one item in my ACBC I used a constructed list so that only 6 of the 18 items are included in the conjoint analysis. When I check the results it shows that also the items that weren’t included in the analysis show very utility scores.
I used the recommended “Inferior to Included Levels” option for HB estimation for levels that have not been included. I checked the results and it seems that one attribute level has an average utility of 132,30 even though no one chose that level to be part of the survey, so it wasn't included in the ACBC. It has the highest utility score of all levels.

It doesn’t really make sense to me that an attribute level that no one chose to be part of the survey has such a high utility.
I also used the setting “Unavailable (Zero Probability of Choice)”  with a low part-worth utility of -10. If I do so, the same attribute level has a utility of -53,78.

Do you have any explanation how that is possible?
asked Feb 28, 2018 by Julia

1 Answer

0 votes
That doesn't sound right.  Please make sure you are using the most recent version of the Lighthouse Studio software.  In this previous year, I think we fixed a bug that was related to dropped levels of an attribute in HB estimation.

It is best to look at utility scores for each respondent.  Check to make sure dropped levels (by that respondent) are inferior to levels that made it into the ACBC survey.

If you have further concerns, please send your data to the technical support team at support@sawtoothsoftware.com.
answered Feb 28, 2018 by Bryan Orme Platinum Sawtooth Software, Inc. (172,690 points)