LABELED CHOICE DESIGN

[d.asioli]

Hello,

I am doing for the first time a labeled choice design where I have 3 products (cow, plant, precision) varying for price (4 levels) only. I plan to use a simultaneous orthogonal design for the pilot study, see code:
Design
;alts = cow, plant, precision, nobuy
;rows = 16
;block = 2
;orth = sim
;model:
U(cow) = b1*PRICE[0,1,2,3] /
U(plant) = b1*PRICE[0,1,2,3] /
U(precision) = b1*PRICE[0,1,2,3] /
U(nobuy) = b0
$
I have tried several designs and I found a good with 16 choice sets splitted into 2 blocks. I got a D-error of: 0.09 and D-optimality of 75%.
I plan to to run a pilot study with the above design, then use the priors and desigining a Bayesian design.

Please, I would like to check of I am on the right direction.

Best regards,
Daniele

[Michiel Bliemer]

The script looks fine. For model estimation you will need to add two more constants because all four alternatives are labelled. For interpretation purposes, I would normalise the constant of nobuy to zero and include constants in all other alternatives, see below where I added a1, a2, and a3. Note that I left out a utility specification for nobuy, which then automatically defaults to 0 utility.

Note that adding these constants will not change your design in this case because it is an orthogonal design, which does not depend on the specification of the utility function.

Code: Select all

Design
;alts = cow, plant, precision, nobuy
;rows = 16
;block = 2
;orth = sim
;model:
U(cow)       = a1 + b1*PRICE[0,1,2,3] /
U(plant)     = a2 + b1*PRICE /
U(precision) = a3 + b1*PRICE
$

Michiel