choice-metrics.com

Hi All,

I am running my design in Ngene and I'm having trouble with the reject function.
What I want to do is put two constraints into my design that rejects all combinations of:
HA=1 and SR =5
HA=1 and PA=5
I have been using the code below it runs fine but when I check to see if the combinations have been rejected I can still find them.

I would be really grateful if anyone could help?

Thanks,
Sio

Design
;alts = alt1, alt2
;rows = 160
;block=20
;eff=(mnl,d)
;alg= mfederov(candidates=combined_d32.csv)
;reject:
alt1.HA=1 and alt2.SR=5,
alt1.SR=5 and alt2.HA=1,
alt1.HA=1 and alt2.PA=5,
alt1.PA=5 and alt2.HA=1
;model:
U(alt1) = b1[0] * A[1,4,7,10] + b2.dummy[0|0|0|0] * MO[1,2,3,4,5] + b3.dummy[0|0|0|0] * PA[1,2,3,4,5] + b4.dummy[0|0|0|0] * HA[1,2,3,4,5] + b5.dummy[0|0|0|0] * IN[1,2,3,4,5] + b6.dummy[0|0|0|0] * SR[1,2,3,4,5] + b7.dummy[0|0|0|0] * LA[1,2,3,4,5]
/
U(alt2) = b1 * A + b2.dummy * MO + b3.dummy * PA + b4.dummy * HA + b5.dummy * IN + b6.dummy * SR + b7.dummy * LA

$

I ran the syntax with ;alg = mfederov, without the candidate set, and those combinations do not appear in the design. Maybe there is an issue with your candidate set, or maybe it does not have the correct format? The format is:
First column: Design/respondent (i.e., 1,1,1,1,1,1)
Second column: Choicetask (i.e., 1,2,3,4,5,6)
Third column: alt1.a
Fourth column: alt.mo
etc.

Alternatively, you can simply impose the constraints directly by removing those choice tasks that violate these constraints in the candidate set.

Michiel

Hi Michiel,

Thanks for getting back to me. I have double checked the format of the candidate set and that seems to be OK. My design includes 3 overlapping attributes could that be the issue?

If not I can always add the constraints into the candidate set before running in Ngene.

Thanks again,
Sio

Overlap or not, constraints should still be imposed properly on the design so I would not know why the reject constraints are not satisfied unless there are too many restrictions to satisfy all (but in that case Ngene should generate a warning that no design could be generated).

I would be interested to see why it is not working properly so if you could email your CSV candidate spreadsheet to contact@choice-metrics.com that would be great, but if you would prefer not to share that file than that is fine and then indeed the way forward is to impose the constraints directly in the candidate set.

Michiel