Have an idea?

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

How to analyze direct-anchored-maxdiff data when the anchor question consists of a subset of items?

How to analyze direct-anchored-maxdiff data when the anchor question consists of a subset of items?

I used the direct anchor method.
The anchor question consists of a subset of items, based on a constructed list.
The constructed list is based on the following logic:
ADD (MXDList, MaxDiffRankAttValue(MXD,  1))
ADD (MXDList, MaxDiffRankAttValue(MXD,  5))
ADD (MXDList, MaxDiffRankAttValue(MXD,  10))
ADD (MXDList, MaxDiffRankAttValue(MXD,  15))
ADD (MXDList, MaxDiffRankAttValue(MXD,  20))
ADD (MXDList, MaxDiffRankAttValue(MXD,  25))

The anchor question was: 'Which of these are important to you?'

How do I analyze this data; including the anchor treshold?

In other words: what logic do I use in the analysis tool?

Thanks!
asked Apr 22, 2020 by flowresulting (420 points)

1 Answer

+1 vote
I like this method of adding a subset of the items according to the scores on the fly to a followup anchor question!  You showed six items in a followup Select-type question, I think you are saying.

So, let's imagine your follow-up question is a Select-type question with respondents able to multi-select from 0 to all 6 of the items to say that they are important.

To analyze the utilities via HB and estimate the anchor score as 0 (and the other scores with respect to the anchor being 0):

Go to Analysis Manager, add an HB run, and then on the settings "gear" icon next to the Analysis Types HB dropdown, click the Anchored Scaling option.  Click "Use Direct Binary Approach".  

Now, you see a list of all your MaxDiff items, with a blank cell to specify some syntax for each.  The syntax you specify is the logical expression that equates to true if each item exceeds the threshold.

Let's say your Select type question where you asked about each of six items was called Q10.  We know that if the respondent saw the first item in your MaxDiff list of items as "checked" then Q10_1=1.  So, for each MaxDiff item, the logic you specify in that table is:

Q10_1=1
Q10_2=1
Q10_3=1
etc.

Even though each respondent saw just six items, you need to specify all 25 items logic in the table, because across all respondents probably each item was shown at least once and you need to have logical syntax to pick that up.

If you used a Grid to show your anchoring question, then you'll have a double-dot notation to refer to each item's checkbox response.  Let's imagine your grid that you used for the Anchor question was named Q11.  We know if the first item in your MaxDiff list of items was "checked" in the grid question then Q11_r1_c1=1.  So, you specify the logic for each cell in your table as:

Q11_r1_c1=1
Q11_r2_c1=1
Q11_r3_c1=1
etc.
answered Apr 22, 2020 by Bryan Orme Platinum Sawtooth Software, Inc. (200,340 points)
Thanks a lot Bryan!
Warning when I use Q10_1=1 logic etc.
...