Have an idea?

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

Max-Diff: excluding an item at the analysis stage

I have a Max-Diff survey where we included 25 items in the exercise.

The data capture and HB analysis has gone very well.

I was wondering what the process was for excluding 1 of the 25 items and re-running the analysis?

Is it possible? If so, what steps do I need to take?

asked Nov 16, 2018 by Paul Moon Platinum (101,255 points)

1 Answer

0 votes
If you want to report the results absent one of the items, it is better just to keep the original HB run, but to remove that item's utility score and adjust the remaining (probability) scores such that they sum to 100%.  

It would add extra noise to the data to try to remove the tasks or remove the comparisons that involved that item and to recompute the scores.
answered Nov 16, 2018 by Bryan Orme Platinum Sawtooth Software, Inc. (201,565 points)
Thanks Bryan.

So to be sure I apply the correct factor to the remaining 24 items in the original HB analysis, would I apply the following ...

Multiple all remaining 24 item scores by: 1/(1-score of removed item)

Regards Paul.
Yes, if you are using the probability-scaled scores that sum to 100%, that's what you do after removing one of the items to scale the remaining back up to sum to 100%  Nice job!
Thanks Bryan. Makes sense. Glad the Sydney event is being put to good use so quickly. Fantastic!