Have an idea?

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

Can I use utilities for a level as a dependent variable in multiple regression?

Let's say that I have 3 levels (red, blue, green) for an attribute (color).  In addition to the conjoint exercise, I also ask respondents questions about red (e.g. 'red is a good value', 'red is feminine', etc).

Could I use the utilities for red as my dependent variable and the respondents ratings about the attributes for red as my independent variable?


Many many thanks!
asked Dec 17, 2014 by Ryan

2 Answers

0 votes
Yes, I think that's possible to do.  Just make sure you're working from normalized part-worth utility data (zero-centered diffs) rather than the raw utilities from HB.  Raw utilities from HB can have big scale factor differences between respondents, depending on how much response error they experienced when answering CBC tasks.
answered Dec 17, 2014 by Bryan Orme Platinum Sawtooth Software, Inc. (128,265 points)
0 votes
One thing to be aware of is that you're likely to see lower R-squared statistics than you may be used to when you're using rating scale data to predict rating scale data.  You model isn't benefiting from "shared method variance" and this will show up in a reduced R-squared.  This doesn't mean your regression is bad, just that it isn't being inflating by using similar measurement methods to predict one another.
answered Dec 17, 2014 by Keith Chrzan Gold Sawtooth Software, Inc. (47,900 points)
...