choice-metrics.com

Hello,

I'm writing to ask a question why the code does not run with * for an alternative.
I have 5 attributes and 2 constraints. I wanted to avoid pairing 0, 1 of attribute B with 2, 3 of attribute A.
I ran the following code and didn't run. So I removed * from altB and it started running immediately. I'd like to know why it's the case and if there's another way to prevent from having attribute level dominance.

Code: Select all

Design
;alts = altA*,altB*, none

;rows = 36
;block = 3

;eff  = (mnl,d)
;alg  = swap(stop=total(600mins))

;cond:
if(altA.B <>2, altA.A = [0,1]),
if(altB.B <>2, altB.A = [0,1])

;model:
U(trtA)=b01[0] +
       b1.dummy[0|0|0]        *A[3,2,1,0] +
       b2.dummy[0.02|0.01]  *B[2,1,0]   +
       b3.dummy[0.02|0.01]  *C[2,1,0]   +
       b4.dummy[0|0]           *D[2,1,0]   +
       b5[-0.01]                     *E[0,1,2,3] /

U(trtB)=b02[0] +
       b1                   *A +
       b2                   *B +
       b3                   *C +
       b4                   *D +
       b5                   *E $


Thank you for your time reviewing.

You only specified non-zero priors for attributes B and C, so Ngene will ignore the other attributes when looking for dominant alternatives. There are only 9 possible combinations of attribute levels B and C, and most of them will have a dominant alternative. Further, you rule out some attribute level combinations for B as well, therefore there are hardly any combinations left while you are asking for 36 rows without any dominant alternatives.

The solution is to set priors for A, D, and E as well. Another solution is to decrease the number of rows.

Note that you need to use altA and altB in your utility specifications instead of trtA and trtB.

Michiel