Have an idea?

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

In a MaxDiff, how do I restrict 1 of 23 items to only be shown to a specific subgroup?

In a MaxDiff I have to set up, one item is supposed to be shown to English as a second language (this is a MaxDiff about TV shows).  There are 23 items in the MaxDiff total.

I looked through the help manual, and it appears that this is possible if I

1. Add a new item for people with English as a first language
2. Set this up as a constructed list with all 24 items (23 + 1 new item)
3. Pick the proper 23 items to show based on which subgroup they are part of.

Then, each subgroup only sees 23 items, and only 22 of those overlap.  

I think I understood that process properly.  Could someone confirm?

If so, then do I just do one HB estimation with all the data and only report the results by subgroup, excluding the items that weren't seen (and aren't relevant)?

Thanks!
asked Nov 1, 2013 by Joel Anderson Bronze (1,310 points)

1 Answer

+1 vote
If I'm understanding, 22 items are common to the two groups.

The 23rd item is only to be displayed to group1

The 24th item is only to be displayed to group2

You can indeed use a constructed list (following the instructions in the SSI Web Manual at: https://www.sawtoothsoftware.com/help/issues/ssiweb/online_help/index.html?hid_web_maxdifflistofitems.htm)

This will result in the 23 appropriate items to be shown to group1 and the 23 to be shown to group2.  Across both groups, 24 items will have been shown.  

When you analyze the results with HB, 24 item scores will be computed for each respondent.  Group1 respondents will get a "draw near the mean for Group2" for the item that only Group2 saw, and vice-versa.  Thus, you'll want to ignore/drop the score for the item that doesn't pertain to each respondent post hoc HB estimation when summarizing the scores.  

The nice thing about this approach is that HB can share information to its benefit across the 22 common items, yet also compute item scores for the 1 unique item to each of the respective groups.
answered Nov 1, 2013 by Bryan Orme Platinum Sawtooth Software, Inc. (144,240 points)
Thanks for confirming and explaining how group1 vs. group2's results will look for the items not seen by their groups.

All makes sense.
MaxDiff : some items filtered for some targets
...