choice-metrics.com

Hello,Professor, I'm William and I'd like to ask a question about D-efficient design:
I currently have four attributes, including three regular attributes and one payment attribute, where the payment attribute is (50, 100, 150, 200 yuan). The specific code is as follows.

design
;alts = alt1*, alt2*,alt3
;rows=12
;block=2
;eff=(mnl,d)
;model:
U(alt1) = b2.dummy[0.0001|0.0002] * b2[1,2,0]
+b3.dummy[0.0001|0.0002]* b3[1,2,0]
+b4.dummy[0.0001|0.0002]* b4[1,2,0]
+ b5[-0.0001]* b5[50,100,150,200]/
U(alt2) = b2.dummy* b2
+b3.dummy* b3
+b4.dummy* b4
+b5 * b5
$
I decided to generate 12 scenarios and then divide them into two parts, but in the scenarios I generated, the willingness-to-pay attributes are always paired as: 200~50 and 100~150. Is this fixed pairing inefficient? How can I make them mix together, like 200~50, 200~100, 200~150, 100~150, and so on?

This is an often asked question here on the forum. Yes that pairing is more efficient if you use (near) zero, but I agree this is often not desirable for other reasons. You have two options:

1. dummy code all your numerical attributes when you are using (near) zero priors - this is what I do
2. use an orthogonal design

Michiel