Have an idea?

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

None utilities in HB


I am doing research to the adoption of a certain product. I have measured how price conscious customers are (using a 5-Likert scale) and have assigned them into three groups (low, neutral and high) accordingly. I would like to know if there is a difference between these groups in terms if they choose the 'none' option (=adoption rate) yes or no.

I did an ANOVA so far to see if there  is a statistical difference between the groups. That means I compared the individual level utitilies from the none option from group 'low' with the utilities in group 'neutral' and 'high'. Now, I doubt if this is legitimate as the magnitude of the none-utility varies among the respondents, so extreme non-adopters could for example mislead the conclusion.

Hope someone can confirm if this is the right approach or what would be a better alternative, thanks!
asked Jan 18 by gakker (140 points)

1 Answer

0 votes
Three options for you:

First, if the outliers in None utility are large then you should have detected that as a violation to sphericity in your ANOVA.  If you pass the test for sphericity, then I think you're OK to go with the ANOVA.

If you're still concerned about outliers, you might try this as a non-parametric test instead.  Say use the 33.3rd and 66.7th percentiles of the None utility distribution to create high, medium and low None utility groups, then cross those groups with the groups from your low/neutral/high groups from your rating scale and run a chi-squared goodness-of-fit test.

Or use a substitute for the None utility that has no chance of getting extreme outliers - simply use the number of times each respondent chooses the None alternative and run that by your three rating scale classes.
answered Jan 18 by Keith Chrzan Platinum Sawtooth Software, Inc. (110,575 points)
Thanks for your swift reply, Keith!