Have an idea?

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

putting constraint on CBC changes attribute importance

Hi, we are working on a CBC dataset and the price attribute turns out to be the most important out of all five attributes. However, the price sensitivity is really low in simulation. I took  a look at the individual utilities and found some respondents have negative utilities on the low price and positive utilities on the high prices. So I imposed "from best to worst" constraint on price attribute.  The new estimation shows price is  almost the lease important attribute of all.

Should I impost constraint in this case? if not, how do I interpret the inversed individual utilities on  price?
asked Aug 16, 2019 by Xiaowen (440 points)
This is a situation where it's difficult to provide advice without digging in to things more, but here's some thoughts:

Importance is just calculated by taking the range of an attribute and dividing it by the sum of all ranges.

If an attribute is important, then that means it has a larger change in utilities from the best to worst level, which means it should have an effect in simulations.

If my utilities were -10 to + 10, but they were reversed, applying a constraint would push them down towards 0 and 0, which would drastically affect the calculated importance.  It would technically "fix" the direction of the attribute, but it doesn't suddenly make your model correct.  It's very concerning that things are going strongly in the wrong direction.

If people had a reversal on price, then you would want to begin to see if there was a programming or communication error in the survey, or if there is any evidence that price is an indicator of quality (think bike locks, I don't want to always buy the cheapest bike lock).  There is probably a larger problem at play here, and applying a constraint probably is not the correct approach.

You can always e-mail in your files to support@sawtoothsoftware.com as well.

Your solution to the original question

Please only use this to answer the original question. Otherwise please use comments.
Your name to display (optional):
Privacy: Your email address will only be used for sending these notifications.
Anti-spam verification:

To avoid this verification in future, please log in or register.