choice-metrics.com

Hi All,

I am designing a labelled CE focused on a wildlife product but Ngene keeps crashing when I run the code below. The priors are largely guesswork based on past literature.

Any thoughts on why this is the case would be most welcome.

Best,
Diogo

Design
;alts= Bear, Synthetic, Porcupine, OptOut
;rows = 24
;block=2
;eff = (mnl,d,mean)
;bdraws = halton(500)
;cond:
if(Bear.Source=0 AND Bear.Form=0, Bear.Price1=[15000,75000]),
if(Bear.Source=0 AND Bear.Form=1, Bear.Price1=[100,1500]),
if(Bear.Source=1 AND Bear.Form=0, Bear.Price1=[8000,12000]),
if(Bear.Source=1 AND Bear.Form=1, Bear.Price1=[10,250]),
if(Porcupine.Source=0, Porcupine.Price3=[350,400]),
if(Porcupine.Source=1, Porcupine.Price3=[200,250])
;model:
U(Bear) = b1[0] + b2[(n,0.03,0.01)]* Source[0,1] + b3[(n,0.02,0.01)]* Form[0,1] + b4[(n,-0.02,0.01)]* Travel[10,30,60,120]+ b5[(n,-0.02,0.01)]* Price1[10,100,250,1500,8000,12000,15000,75000] /
U(Synthetic) = b6[0] + b4 * Travel[10,30,60,120]+ b7 [(n,-0.02,0.01)]* Price2[20,110,200] /
U(Porcupine) = b8[0] + b2 * Source[0,1] + b4 * Travel[10,30,60,120]+ b9 [(n,-0.02,0.01)]* Price3[200,250,350,400] $

Ngene does not crash in my case, it simply gives a warning that no feasible design could be found. The issue seems to be in your priors, they should always be chosen with great care, preferably through a pilot study. A prior of -0.02 is extremely large for an attribute with a level 75000 since this would mean a contribution to utility of 375, which results in a completely dominant attribute price1.

I suggest you set all your priors to zero unless you have parameter estimates from a pilot study.

Best wishes,
Michiel