Do the screener questions in the ACBC use a multinomial-logit setup?

Question

Thanks for this information, Bryan. I'm also curious as to how Sawtooth actually handles the responses. For example, if we use the screener questions and change the choice options from a "choose/not choose" format to a "choice 1/choice 2" format, I wonder how this would affect the analysis. If Sawtooth uses a multinomial-logit setup, then all options are identified as a bundle, and it should not matter if the option is Not Choose or Choice 2 since this model decomposes the impact of the characteristics of the person making the choice or receiving treatment, not the characteristics of the alternatives presented. Do the screener questions in the ACBC use such a multinomial-logit setup?

Answer 1

The tasks from the screener question of ACBC are coded for MNL estimation as binary logit, where category two of the dependent variable is the "None" alternative (the outside good).  Analysts can decide to do what they please with that extra "none" parameter...either use it or ignore it during analysis.