Have an idea?

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

How does Dual-Response-None work?


I consider doing a dual-response-none CBC-HB analysis and want to compare results for both cases, if the respondents do have the none option and if they don't. How would I proceed?

More specifically, is it correct to do the estimation twice, one time without having ticked the "include none parameter if available" and a second time with this option ticked? Or would such a comparison be done differently?
asked Sep 1, 2015 by Ulf

1 Answer

0 votes
Best answer
Yes, I think unchecking that box "include none parameter" should do it.  In the "build report" at the beginning of the estimation process it will report to you how many choice tasks it is about to use in the estimation.  This will help you verify that the HB estimation is only going to used the "forced choice" where the None alternative as not available.

Of course, unchecking that box doesn't change the psychological situation facing respondents that they learned that the None question always followed the forced choice.  So, you cannot with your data set ever estimate exactly what would have happened had you not included a dual-response None task in the questionnaire.
answered Sep 1, 2015 by Bryan Orme Platinum Sawtooth Software, Inc. (128,265 points)
How would you go about subsequent simulations in SMRT? Would it be correct to export the respective estimates (with and without none-parameter) as separate files from SSI into SMRT and run two identical simulations with them, setting the none weight to 0 or 1 respectively?

Another doubt: I get a warning when using dual none responses that "newer versions" of SMRT can import such data but ignores the nones. I am using SSI 8.4.6 and SMRT 4.23. Can I analyze dual responses with these versions or do I have to get another one?
I think you'll need to import the .HBU files separately into two different SMRT projects (one .HBU has a None parameter and the other does not).

The warning is only regarding if you try to read a .CHO file into SMRT for the purpose of running counts or aggregate logit analysis (newer versions of SMRT are able to, but drop the None information from the Counting and logit analysis).  Most people probably just import the .HBU file via Analysis + Run Manager + Add within SMRT to run market simulations and they never bother to import the raw choice data (the .cho files) into SMRT for the purpose of running Counting analysis and logit.