Hi Jules,
I wasn't familiar with the "dual response none" approach, so I am basing my response on the description provided in
this website. It basically state that -in the context of a stated preferences survey- the method consists in splitting each choice exercise in two stages: a forced choice between alternatives, followed by a binary choice between the selected alternative and an opt-out.
The webpage, however, does not state how the estimation is performed, making it difficult to reproduce the process in Apollo. On the other hand, Brazell et al. (2006) state that the opt-out alternative fulfils the independence of irrelevant alternatives (IIA) property. If so, estimating a model for only the first stage choice would yield consistent estimates for the attribute preference parameters (beta). If we also wanted an alternative specific constant for the opt-out alternative, we could model both stages as a single one with all the alternatives (including the opt-out) at the same time.
Again, I do not know what approach Sawtooth Lighthouse Studio uses, but both approaches can be estimated in Apollo.
- To estimate ignoring the opt-out in Apollo: Simply use an MNL considering only the first stage choices (the forced choice).
- To estimate the ASC of the opt-out in Apollo: Mix both stages in a single choice with all alternatives, and define the utility of the opt-out as a single constant (ASC). Estimate using a regular MNL model.
Sorry for the lack of specific in my answer, but without knowing exactly what Sawtooth does, we can hardly reproduce it in Apollo.
References:
Brazell, J.D.; Diener, C.G.; Karniouchina, E.; Moore, W.L.; Séverin, V. and Uldry, P-F. (2006) The no-choice option and dual response choice designs. Marketing Letters 17, 255 - 268.