Have an idea?

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

Usage of Dual-Response-None

Dear community,

I am thinking about using the dual response none in my CBC experiment. Therefore, I have some open questions:
1. Is there any difference in defining and analysing holdout tasks when using dual response none in my CBC?
2. Can I get seperate estimates for responses of the first stage (forced choices) without directly being combined with their responses from second stage (yes/no choices)? The standard estimates are combined from responses from first and second stage, arent´t they? Is there any evidence that pooling this data in one data set violates the IID assumption? Is there any documentation available how exactly LHS is jointly extimating the responses from the two stages?
3. I assume that WTP can be estimated as with CBC experiments containing "normal" none option, right?

Any advice is highly appreciated!
asked Apr 7 by anonymous

1 Answer

0 votes
1.  Not sure what kind of difference you're talking about.  You can still ask and model holdout choices.  

2.  Yes, you can estimate them separately.  And good point about pooling and IIA - I have seen some papers discussing IIA when the none alternative draws disproportionately from the other alternatives (though if I remember correctly, it was in the context of labeled alternatives)

3.  Yes, right
answered Apr 7 by Keith Chrzan Platinum Sawtooth Software, Inc. (102,150 points)
Hi Keith,

thank you very much for your quick and helpful response!

One more questions on this topic:
Is there any documentation available how exactly LHS is jointly extimating the responses from the two stages? I just found the information that both stages are estimated jointly but not how exactly. I might need some more details for the theoretical part my research.
Thank you in advance.
Yes, that's covered in the Coding chapter of our Becoming and Expert in Conjoint Analysis book.
...