Separate HB runs and attribute importance interpretation

Question

Dear Sawtooth Community,

Here I have two questions about ACBC.

If I want to know the relative importance of attributes and utility values for two groups respectively, i.e., laypeople vs. professional in my work, should I choose the approach of running separate HBs (one HB only includes laypeople participants and another HB only includes professional participants), or combine the two groups as an entire sample and run HB for the entire sample and then segment based on if the participant is a professional? I know the result from the two approaches is different.

Another question is about the interpretation of attribute importance. If the importance value of the attribute A that was ranked in the first place was not statistically significantly higher than the second ranked attribute B (Wilcoxon test result from SPSS), do you recommend interpreting attribute A was the most important attribute in comparison to others? Or is it more appropriate to say attributes A and B were the most important attributes in comparison to others?

Thanks in advance.

Answer 1

For the first question, if I have two segments and it makes sense the two would have different utilities, I usually run a single analysis but I'd use the segment variable as a covariate in my model.  Our tests have shown that this does pretty much what doing two separate runs would do in terms of allowing the utilities to be different, but it allows you to do the two at once.   

For the second question, why would we EVER say two things were different if statistically they were not?  I don't know your reason for running Wilcoxan (which is a non-parametric test) instead of a repeated measures ANOVA, but if you're going to go to the trouble of running statistical tests, I think you should let "not significantly different" mean something.  And it doesn't mean any more or less because you're doing conjoint analysis than it does anywhere else.

Comments

Thanks for your answers, Keith!