choice-metrics.com

Dear Moderators,
I compared the resulting form three different syntaxes about prior for neither choice. The D-error values the highest is a negative sign, no prior and positive sign, respectively. Could I choose the model from D-error value?

Code: Select all

design
;alts=alt1*,alt2*,alt3
;rows=12
;eff=(mnl,d)
;model:
U(alt1)=b1.dummy[0.000001]*format[1,0]+b2.dummy[0.000001]*con_ser[1,0]+b3[-0.000001]*time[10,20,30,40]+b4.dummy[0.000001|0.000002]*area[1,2,0]+b5[-.000001]*cost[0,100,200,300]/
U(alt2)=b1*format+b2*con_ser+b3*time+b4*area+b5*cost/
U(alt3)=b0[-0000001] or b0[0] or b0[0.0000001]
$


Sincerley yours
Sukunta Muadthong

You cannot select the model based on the D-error as D-errors are not comparable across different models and priors.

In your case, using -0.000001 or 0 or 0.000001 lead to the same result and the sign is ignored in the dominance checks, therefore I suggest simply using the value of 0. Note that you get a different value for -0000001 since this is -1, not -0.000001.

Michiel

Dear Michiel Bliemier,
Thank you so much. That is my typo of value 000000001, I mean 0.000001 and I understand that the utility of neither choice was assumed to zero in Ngene. Then we do not add the prior for neither choice.

Sincerely yours,
Sukunta