After inserting priors,MNL model failing to converge

[JasonOng]

Dear NGENE team

After collecting information from my pilot, I tried inserting priors into my D-efficient MNL design but NGENE seems to be having trouble finding a design
I included coefficients that were statistically significant and left the rest as 0
what do you suggest I do?
thanks for your help

Kind regards,
Jason Ong

Code: Select all

Design
; alts = A*, B*, C
; rows = 24
; eff = (mnl,d) 
?; bdraws = gauss(5)
; block =2
; con
; model:
U(A) = b1.effects[0|0|-0.9]*Cost[500,1000,2000,0] +
       b2.effects[0|0|0|-1|0|0|0]*Loc[1,2,3,4,5,6,7,0] +
       b3.effects[0|0]*Test[1,2,0] +
       b4.effects[0|0|0]*Person[1,2,3,0] +
       b5.effects[0]*Access[1,0]/
U(B) = b1*Cost +
       b2.effects*Loc +
       b3.effects*Test +
       b4.effects*Person +
       b5.effects*Access/
U(C) = b6[-1.4]
$

[Michiel Bliemer]

Checking for dominance (you added a * to alts A and B) with mostly zero priors is problematic since every choice task will have a dominant alternative. You either need to remove the asterisks or you need to insert non-zero priors.

Instead of using zeros for non-significant priors, you can use normally distributed Bayesian priors where the parameter estimate is used as the mean and the standard error is used as the standard deviation. Note that you have a lot of priors so you will need to assume some priors to be fixed instead of random. Using gauss(5) will do millions of draws if you use a lot of Bayesian priors, so you may need to reduce this to gauss(2) or gauss(3).

Michiel