Constraints within a block?

[mikp]

Hi all,
First off thank you providing this great forum to support usage of NGENE toolset.
I am developng a design for a SP study on the recovery of a threatened species.

There are three alternatives:
Status Quo, Option 1, Option 2

and three attributes for each alternative:
- population number
- time to recovery
- cost associated with option

Population number has 2 possible levels
Time to recovery attribute has 3 possible levels
Cost has 6 possible levels

Each respondent participating in the survey will be asked to complete three choice situations with varying attribute levels.

I am generating an WTP design with the following code:

Code: Select all

Design

;alts = alt1*, alt2*, SQ

;rows = 36 
;block=12
;eff = (mnl, wtp(ref1))  ;wtp = ref1(b1,b2/b3)
;cond:
if(alt1.fish >= alt2.fish and alt1.time <= alt2.time, alt1.cost > alt2.cost),
if(alt2.fish >= alt1.fish and alt2.time <= alt1.time, alt2.cost > alt1.cost)

;model:

U(alt1) =  b1[0.0001]*fish[40000,70000] + b2[-0.04]*time[15,25,50] + b3[-0.01]*cost[40,75,120,200,250,300] /
U(alt2) =  b1 * fish +  b2 * time  + b3 * cost /
U(SQ) = b4[0]
$


Note: the constraints need to included so that cost rises with increasing fish populations and/or decreased time to recovery.

I have the following questions:

1. Can I add additional contraints in NGENE so that within any given block, the respondent does not see the same combination of attribute levels at different prices. A colleague warned me that some respondents may get confused when the price varies for the same combination of attributes across choice scenarios. I'm not sure this will really be a problem, but I do remember a few of our focus group participants asking similar questions.

2. If it is not algorithmically feasible to have constraints within blocks, I may try to change levels manually.
Is this a bad idea? I realize this is not an NGENE question, but any guidance on this issue would be greatly appreciated.
If I do change them manually, I assume I can check the new design using the eval command.

thanks for any input.

[Michiel Bliemer]

1. Constraints can only be imposed within individual choice tasks, not across choice tasks (in a block). So the answer is no. However, in an stated choice experiment, choice tasks are hypothetical of a "what-if" charachter. So asking "would you choose A or B if the price is $10?" and then followed by "now consider a different scenario, would you choose A or B if the price is $20?" should in my view not be problematic, as this is the nature of stated choice experiments with multiple varying scenarios. Each scenario is independent of the previous, and this you can make clear in the survey.

2. There is no problem changing them manually, it is better to have a design that you are happy with, then a design that is on paper (theoretically) more efficient. I would rather have less efficiency if I can make the scenarios more realistic (which you essentially do with constraints, either using Ngene or manually as you indicate). Indeed you can always check your design using the eval command. So do not shy away from making changing a-posteriori if you believe it makes the design more realistic or easier to understand. It is not an exact science.