Choice set consisting of 3 health states

[CMA]

To deepen my understanding, I would like to ask a question.
I am planning experiments with two health states (state A, state B) (= part A) and three health states (state A, state B, death) (= part B).
For part B, I want to design it so that state A and state B are the same as in part A, but with the addition of death.
However, I don't have a plan for this.
Should I avoid using Ngene to design an experiment with two health states (state A, state B) and then retroactively add "death" to create an experiment with three health states (state A, state B, death)?

[Michiel Bliemer]

You can create a single efficient design for estimating a joint model, I suggest that you read the Ngene manual and look at the ;fisher property, or at the model averaging section 7.4 in the manual.

The script could look something like:

;alts(model1) = A*, B*
;alts(model2) = A*, B*, death

;eff = fish(mnl,d)

;fisher(fish) = design1(model1[0.5],model2[0.5])

;model(model1):
U(A) = ... /
U(B) = ...

;model(model2):
U(A) = ... /
U(B) = ... /
U(death) = b0


Of course you could also generate an orthogonal design using ;orth = ood for health states A and B and then add the death alternative later on. This does not necessarily optimise the design for the data collection, but it will still be a useful design nevertheless.

Michiel