choice-metrics.com

Hi,

I have a question to the board about developing a design with constraints. I am studying individuals’ preferences for coastal ecosystems under sea level rise scenarios. In our scenarios, the type of management intervention should make a difference in the potential outcomes. For example, one type of engineering intervention will change freshwater flow into the coastal ecosystem and may also influence aesthetics. Here is an example of my code:
Design
;alts = alt1, alt2, alt3
;rows = 20
;block =4
;eff = (mnl,d)
;alg = mfederov(candidates = 1000)
;reject:
alt1.eng = 0 and alt1.salt = -12,
alt1.eng = 1 and alt1.salt = 12,
alt1.eng = 0 and alt1.fresh = 18,
alt1.eng = 1 and alt1.fresh = 0,
alt2.eng = 0 and alt2.salt = -12,
alt2.eng = 1 and alt2.salt = 12,
alt2.eng = 0 and alt2.fresh = 18,
alt2.eng = 1 and alt2.fresh = 0
;model:
U(alt1) = b0[-.5] + b1[.1]*salt[-12,-6,6,12] + b2[.125]*fresh[0,6,12,18] + b3[.075]*upland[1,2,3,4] + b4.effects*eng[0,1] (10,10) + b5[-0.125]*fee[1.5,2.5,5,10,15]/
U(alt2) = b0 + b1*salt+ b2*fresh + b3*upland+ b4.effects*eng + b5*fee $


In this scenario, the type of management intervention (eng), influences the potential outcomes of two attributes (salt & fresh). Is it problematic to induce this type of correlation into the design? It makes sense from a real world perspective, but I am looking for a good example in the literature. At this point, we are trying to develop a realistic design that will be used in our pilot study. Our pilot will then inform a redesign.

Putting constraints on a design decreases the statistically efficiency, but increases realism. We typically argue that realism is more important than efficiency. So in your case, the constraints seem to be added because of realism and this is likely a good idea. Many people (us included) use such constraints to improve realism.