Have an idea?

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

Can I delete two attributes from MaxDiff and CBC after the study is done

We have a study with 30 attributes in a MaxDiff and 30 attributes with 2 to 5 levels in a PPCBC.  Now, after the study is done, my client wants to throw out 2 attributes.  Mathematically we can throw out attributes, re-scale scores in MaxDiff.  But I do not believe it is correct.  

Both attributes are highly correlated with several other attributes.

But what happens with the PPCBC?  One attribute that the client wants to throw out, price, was shown in every exercise.  The other attribute, quality, rotated across 25 PPCBC exercise.

Please provide your expert opinion.

Thank you!
asked Jun 28, 2019 by anonymous

1 Answer

0 votes
You can toss them AFTER you estimate utilities.  I would not toss them before estimating utilities because the variability that was explained by the to-be-tossed attributes will want to avoid being unexplained variability so it will likely attach itself to something.  This might not be a problem if all your attributes were uncorrelated at the respondent level, but with that many attributes, you likely have different attributes correlated with the to-be-deleted ones for different respondents.  

As a result I think it will be cleaner to punt them AFTER utility estimation.
answered Jun 29, 2019 by Keith Chrzan Platinum Sawtooth Software, Inc. (83,450 points)
That is exactly what I did.  Thank you.