Have an idea?

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

Setting up Express Max-Diff


I am new to SSI Web. I am currently using the version 8.4.8.

I am trying to setup an express max-diff question in SSI. I have read about and tried the following approaches:

1. Using a constructed list for the setup - This seems doable as I was able to set up the problem by creating a parent list of all attributes and creating a constructed list by randomizing and then selecting a certain number of attributes from this randomized list. Is there any other setting that should be taken care of while setting up an express max-diff in this way?

2. Using the method outlined in the Wirth/Wolfrath paper from the Sawtooth Conference 2012 - Though I believe I understand the concept used in this method, but I am having some difficulty trying to understand how to efficiently and correctly implement it in SSI. I especially cannot understand the meaning of the following step mentioned in the paper, "Before fielding the study using SSI Web, the item numbers in the questionnaire versions (which, in our example always range from 1 to 30 in all tasks) must be replaced by the actual item numbers (which, in our example, range from 1 to 100). This “translation” can be done using standard software, such as R." What exactly is the “translation” being talked about here? Can someone illustrate and help me understand it a bit more clearly?

Thanks in advance for any help!
asked Sep 3, 2018 by prashant (180 points)

1 Answer

0 votes
Express MaxDiff randomly selects a subset of a larger list of items to show each respondent.  The hope is to show each of the randomly selected items two or more times per each respondent.  It has weaknesses relative to sparse MaxDiff (where respondents see all items and each item is shown 1x per respondent).  Based on numerous tests we've conducted, we tend to prefer sparse MaxDiff or Bandit MaxDiff when dealing with a large list of items.

Now, on to your question.  Wirth/Wolfrath were not using our software to do Express MaxDiff, so it required extra data processing and recoding (translation) steps as they mentioned.  Our software handles these recoding steps automatically.
answered Sep 3, 2018 by Bryan Orme Platinum Sawtooth Software, Inc. (200,115 points)
Thanks for the response, Bryan.

Is it possible for you to briefly tell me how to carry out the procedure in SSI Web? I am able to generate the design blocks for different respondent groups. What I am not able to figure out is how to assign these blocks to different respondents and create different max-diff sets for them.
This all happens automatically.

For example:

1.  Decide how many items to draw for each respondent (say, 30).  Create a constructed list where you randomize the full list of items and then delete all but 30 of them by using the SetListLength instruction to limit the length of the list to 30 items.

2.  Design a MaxDiff study using 30 items.  Use the defaults to create 300 versions of that questionnaire.  Point the MaxDiff exercise to use the constructed list (which is built to have 30 of the items per respondent), NOT the pre-defined list.  

3.  When you field the study, it will assign respondents to different versions.

All this is described in the Help for MaxDiff at: http://www.sawtoothsoftware.com/help/lighthouse-studio/manual/index.html?hid_web_maxdifflistofitems.html

Down that document, look for the section entitled, "Using Constructed Lists in MaxDiff Experiments".
Hi Bryan,

Thanks again for the reply. The approach you outlined is same as what I had thought of! I was just wondering if there's any other approach as well or any special settings I should take care of. Thanks for clarifying once again.