Page 1 of 1

Question on efficient design

PostPosted: Fri Jul 23, 2021 3:16 am
by bpaudel
Hi,
I want to use D-efficient design to generate choice sets for my choice experiment. I have generated 4 blocks with a total of 40 choice sets and included 3 alternatives (Option A, Option B and Neither) in a choice set.
Regarding the attributes and levels, I have 4 attributes (Price, CF, HR, T) with their levels mentioned below:
Price($/lb): 5, 7.75, 10.5, 13.25, 16
CF: Yes, No
HR: Yes, No
T: B, No label

Dummies in design:
CF (Yes)=1, CF (No)=0
HR (Yes)=1, HR (No)=0
B=1, No label=0

We want price to be continuous and fixed while other attributes are random. I have a following ngene design.

Design
Design
;alts= optA*, optB*, Neither
;rows=40
;eff=(mnl,d, mean)
;block=4
;alg= mfederov(candidates = 2000)
;bdraws=gauss(4)
;model:
U(optA) = b1 [-0.367] * Price[5, 7.75, 10.5, 13.25, 16](6-10,6-10,6-10,6-10, 6-10)
+ b2.dummy[(n, 0.7898, 0.1089)]* CF[1,0]
+ b3.dummy[(n, 0.1193, 0.022)]* HR[1, 0]
+ b4.dummy[(n, 0.3954, 0.1089)] * T[1, 0] /
U(optB) = b1 * Price
+ b2 * CF
+ b3 * HR
+ b4 * T /
U(Neither) = b0 [(n, -3.5339, 0.566)]
$
Are there any issues in the design? If any, please let me know how can I improve the design?

With regards,
bpaudel

Re: Question on efficient design

PostPosted: Fri Jul 23, 2021 10:09 am
by Michiel Bliemer
That looks good to me.

I am wondering why price has a fixed prior given that it the attribute with the largest impact on utility and therefore the design will be most sensitive to the value of this prior and would benefit from a Bayesian prior.

Michiel