choice-metrics.com

Dear all,
I haven't use ngene for a while and perhaps I am a little rusty, I have the manual and all sources of help but even after minimizing the design to 3 alts. (I had 5 to start with) and skipping the random draws, it still doesn't return a design. The priors based on a pilot survey analyzed with MNL

Code: Select all

design
design
;alts = money, shopping_list,optout
;rows = 24
;block=8
;eff = (mnl,d)
;model:
U(money) =b0[-.186]+b1[-.001]*cost[1,2]/
U(shopping_list) =c0[.156]+c1[-.2285833]*comitment[1,2]$

I keep getting the error msg.

A valid initial random design could not be generated after approximately 10 seconds. In this time, of the 94199 attempts made, there were 94199 row repetitions, 0 alternative repetitions, and 0 cases of dominance. There are a number of possible causes for this, including the specification of too many constraints, not having enough attributes or attribute levels for the number of rows required, and the use of too many scenario attributes. A design may yet be found, and the search will continue for 10 minutes. Alternatively, you can stop the run and alter the syntax.

Any advice what might be wrong?
Thanks!
Anat

You only specify two attributes each with only 2 levels, so in total there are only 4 possible choice tasks. You are asking Ngene to generate 24, but only 4 exist. If these are the design dimensions that you want, then simply ask all 4 questions to each respondent.

Thank you very much Michiel,
Makes perfect sense.
By the way the equivalent orthogonal design (which used to obtained the priors) did converge with 24 rows (where 4^2=16 seems to be the max. possible choice tasks. What do I'm missing here?
Anat

Code: Select all

Design
;alts = list, freqency,variety, money, optout
;rows = 24
;orth=sim
;block=4
;model:
U(list) = b0+b1[-0.01]*commited[1,2]/
U(variety ) = c1[0.1]*variety[1,2] /
U(freqency) = d1+ b2[0.1]* freqency[1,2]/
U(money)= e1+ b3[-0.1]*cost[1,2]$


The orthogonal design will give you a design, but again it would not be a design that I would use since it again contains replications of some of the same choice tasks. I agree that it is somewhat inconsistent for Ngene to generate an orthogonal design and not an efficient design. Either Ngene should generate both (with replications), or neither.

You are asking Ngene to generate a fractional factorial design (i.e. take a fraction of the full factorial, which in your last case contains 16 choice tasks) either using an orthogonal design or an efficient design. But since the full factorial is only 16, it is a bit odd to ask Ngene to select 24 out of 16.

I guess this is a case where we did not anticipate the user to provide this input, but I believe we should generate an error message in the future to alert the user that this is not a sensible thing to do.

Michiel

Thank you Michiel,
As always, your answers are detailed and very helpful.
There used to be more attributes in this design, and when I removed them eventually, I forgot to adjust the other specifications (I guess I sometimes work on "automatic pilot").
Will bear it in mind next time.
Thanks again!
Anat