Design for binary choice DCE

[oxecon]

Hi there,

I have two related questions. I am designing a DCE where there is just one choice option and which I hope to analyse using random effects versions of a binary logit or binary probit model.
I am considering using the code below:
Design
;alts = alt1, none
;rows = 15
;eff = (mnl,d)
;cond:
If(alt1.A = 1, alt1.B = [1,2])
;model:
U(alt1) = b1.effects[0|0] * A[1,2,3] + b2.effects[0|0] * B[1,2,3] + b3 * C[0.3,0.1,0.01] + b4 * D[0.3,0.1,0.01] + b5.effects[0|0] * E[1,2,3] + b6.effects[0|0] * F[1,2,3]

My two questions are:
Q1: There doesn’t seem to be a tailor made option in Ngene for optimising designs for a binary logit/probit. In this case, by forcing Ngene to do a design with just one alternative, and asking to optimise for MNL, will it essentially be optimising for a binary logit with this code?

Q2: Have I specified the utility model correctly, or should I instead add a constant at the start, as below?
U(alt1) = b1 + b2.effects[0|0] * A[1,2,3] + b3.effects[0|0] * B[1,2,3] + b4 * C[0.3,0.1,0.01] + b5 * D[0.3,0.1,0.01] + b6.effects[0|0] * E[1,2,3] + b7.effects[0|0] * F[1,2,3]

Any help warmly appreciated!

Many thanks,

Stan

[Michiel Bliemer]

Q1: A binary logit model is a special case of the multinomial logit model when there are exactly two alternatives; there is no other difference between a binary logit model and a multinomial logit model. Therefore Ngene will optimise for a binary logit model with this syntax. Ngene cannot optimise for probit models.

Q2: Since you have a labelled experiment by adding the no choice alternative, it is appropriate to add a constant.

Michiel

[oxecon]

Thanks a lot for your help Michiel, much appreciated!