First of all - congratulations on creating such an accessible way of entering into the murky world of DCE design, it has been an enjoyable process thus far.
I am having a little trouble constraining attributes in my syntax below. Our experiment is looking at 5 different products plus an opt out, and we have decided on an unlabelled design after piloting suggested there was too much information in a labelled design. As such, we now want to constrain the frequency with which different products can be used (e.g. every day, 3 per week, weekly, annually, bi-annually) according to the product displayed in each task.
- Code: Select all
Design
;alts=A,B,C,none
;rows=15
;eff=(MNL,d)
;cond:
if(A.product= 1, A.freq = [0,2,3]),
if(B.product= 1, B.freq = [0,2,3]),
if(C.product= 1, C.freq = [0,2,3]),
if(A.product= 2, A.freq = [1]),
if(B.product= 2, B.freq = [1]),
if(C.product= 2, C.freq = [1]),
if(A.product= 3, A.freq = [0,1]),
if(B.product= 3, B.freq = [0,1]),
if(C.product= 3, C.freq = [0,1]),
if(A.product= 4, A.freq = [4,5]),
if(B.product= 4, B.freq = [4,5]),
if(C.product= 4, C.freq = [4,5]),
if(A.product= 5, A.freq = [4,5,6]),
if(B.product= 5, B.freq = [4,5,6]),
if(C.product= 5, C.freq = [4,5,6])
;model:
U(A)= product.effects[0|0|0|0]*product[1,2,3,4,5]+
G[0]*G[55,75,95]+
H.effects[0]*H[0,1] +
freq.effects[0|0|0|0|0|0]*freq[0,1,2,3,4,5,6]+
I.effects[0]*I[0,1]+
J[0|0].effects*J[0,1,2]/
U(B)= product.effects[0|0|0|0]*product[1,2,3,4,5]+
G[0]*G[55,75,95]+
H.effects[0]*H[0,1] +
freq.effects[0|0|0|0|0|0]*freq[0,1,2,3,4,5,6]+
I.effects[0]*I[0,1]+
J[0|0].effects*J[0,1,2]/
U(C)= product.effects[0|0|0|0]*product[1,2,3,4,5]+
G[0]*G[55,75,95]+
H.effects[0]*H[0,1] +
freq.effects[0|0|0|0|0|0]*freq[0,1,2,3,4,5,6]+
I.effects[0]*I[0,1]+
J[0|0].effects*J[0,1,2]/
U(none)=b1
$
Running the above produces the following message:
"Warning: No valid design has been found after 1000 evaluations. There may be a problem with the specification of the design. A common problem is that the choice probabilities are too extreme (close to 1 and 0), perhaps because some or all of the prior values are too large. Also, it is generally a good idea to start with a simple design (MNL, non-Bayesian), then add complexity. If you press stop, a design will be reported, which may assist in diagnosing the problem."
Removing the constraints on one product results in a design running, but this would mean unrealistic choice scenarios. I have played around adding and removing constraints as much as possible but have been unable to generate a design so far.
Any advice would be gratefully received
Matt