Have an idea?

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

Keeping one Alternative constant

I am currently trying to design a choice experiment. This choice experiment requires that one of the options (Alternative 1) is kept constant as a current situation. Just the alternative option (Alternative 2) should be varied over the different attribute-levels. How can I apply this in Sawthoot?

And if possible that the Attributes-level for Alternative 1 are kept constant, is it possible to let one of the Attributes vary in levels?

Thank you in advance.

Best regards
asked Apr 25, 2016 by Jakob

1 Answer

+1 vote
I've done this a few ways in the past.  

One way is just to build the experiment as a one-concept CBC with a standard None alternative.  Then simply type the constant attribute levels into the None alternative.  Unfortuantely in your case, it is Alternative 2 that will be constant in this arrangement.  

The other way is a bit more complicated.  Create the experiment  with two alternatives.  Then export the design and manually code the first alternative always to have your fixed levels.  Then import the design back into the Lighthouse or SSI Web software

Better still would be to build this as a once concept CBC with a standard none.  Then export the design and modify it in Excel:  make the one concept be concept 2 in each choice set and add a concept to be the fixed concept 1.  Again, import into the Lighthouse or SSI Web software and you're ready to go.  This may give you a slightly better design than the second approach above.   

If you run into trouble, our custom analytics group can assist you for a fee:  just contact us at analytics@sawtoothsoftware.com.
answered Apr 25, 2016 by Keith Chrzan Platinum Sawtooth Software, Inc. (65,925 points)
Keith gives good suggestions and he is a power user of the software.  Keith's second suggestion will work just fine for generating the experimental design and showing the tasks to the respondent for proper data collection.  But, you'll need to do an additional step to allow you to estimate the utilities properly for the varying levels vs. the fixed alternative.  You'd need to recode the data after it has been collected similar to the way alternative-specific designs are done in our CBC software.  There are various ways to do that, but all involve some manual data file manipulation.

Another approach that is the formerly "supported" way to do it in the software, though it sometimes doesn't provide the layout for each attribute in each row of the grid as nicely as Keith's suggestion number 2 is to design, field, and compute utilities for this as an alternative-specific design in CBC.  Specifying the full concept text for the fixed alternative is actually done as specifying a single level for the "primary" attribute that looks to the respondent like it's multiple levels formatted across multiple rows (this level appears with no other varying levels of other attributes in the alt-specific design).  The trick is to get it to line up nicely with the other varying concepts.

I'm sure our tech support group deals with this type of question all the time and they are available to you by calling 801 477 4700.