You have specified that all alternatives are unlabelled, while in reality only alt1 and alt2 are unlabelled and alt3 is labelled (no choice altervatives clearly have a different utility function, namely equal to zero).
Therefore, you should use:
;alts = alt1*, alt2*, alt3
generating priors
Moderators: Andrew Collins, Michiel Bliemer, johnr
-
kodhis2000
- Posts: 1
- Joined: Sun Sep 27, 2015 7:26 pm
Re: generating priors
Hi
I have a similar problem in generating my design, please help me identify what could be the problem. i am using the priors from my pilot study in which am trying to value a range of forest ecosystem service.
below is my syntax
?MNL BAYES PRIORS
Design
;alts = altA*,altB*,altC
;rows = 30
;block = 6
;eff = (mnl,d)
;model:
U(altA) = b1[n,-0.45,0.16] * wild[0,1,2] + b2[n,0.68,0.15] * tree[0,1,2] + b3[n,0.05,0.14] * water[0,1,2] + b4[n,-0.04,0.11] * flood[1,2,3] + b5[n,-0.33,0.06] * cost[1744,2683,2951] /
U(altB) = b1 * wild + b2 * tree + b3 * water + b4 * flood + b5 * cost$
i get the following error on running the command,
"no valid design has been found after 1000 evaluations. there may be a problem with the specification of the design. A common problem is that the choice probabilities are too extreme(close to 1 and 0), perhaps because some or all prior values are too large.
Regards
Boscow
I have a similar problem in generating my design, please help me identify what could be the problem. i am using the priors from my pilot study in which am trying to value a range of forest ecosystem service.
below is my syntax
?MNL BAYES PRIORS
Design
;alts = altA*,altB*,altC
;rows = 30
;block = 6
;eff = (mnl,d)
;model:
U(altA) = b1[n,-0.45,0.16] * wild[0,1,2] + b2[n,0.68,0.15] * tree[0,1,2] + b3[n,0.05,0.14] * water[0,1,2] + b4[n,-0.04,0.11] * flood[1,2,3] + b5[n,-0.33,0.06] * cost[1744,2683,2951] /
U(altB) = b1 * wild + b2 * tree + b3 * water + b4 * flood + b5 * cost$
i get the following error on running the command,
"no valid design has been found after 1000 evaluations. there may be a problem with the specification of the design. A common problem is that the choice probabilities are too extreme(close to 1 and 0), perhaps because some or all prior values are too large.
Regards
Boscow
-
Michiel Bliemer
- Posts: 2057
- Joined: Tue Mar 31, 2009 4:13 pm
Re: generating priors
There are making multiple mistakes, but all related to your priors.
1. You have not specified Bayesian priors, but rather you have specified random coefficients in a mixed logit model. You should use b1[(n,-0.45,0.16)] with round brackets around the priors, please refer to the Ngene manual
2. You have not specified a Bayesian efficiency measure nor which Bayesian draws to use, so please add 'mean' or 'median' to your efficiency measure and set the ;bdraws property. I refer again to the manual.
3. Your priors do not seem to make sense. You may want to use priors that come out of a pilot study or other estimations. Looking at your prior for cost, your cost attribute is very large (1744 and larger) and you are multiplying with a large prior (-0.33), this gives a huge utilities. Please use priors that make behavioural sense.
1. You have not specified Bayesian priors, but rather you have specified random coefficients in a mixed logit model. You should use b1[(n,-0.45,0.16)] with round brackets around the priors, please refer to the Ngene manual
2. You have not specified a Bayesian efficiency measure nor which Bayesian draws to use, so please add 'mean' or 'median' to your efficiency measure and set the ;bdraws property. I refer again to the manual.
3. Your priors do not seem to make sense. You may want to use priors that come out of a pilot study or other estimations. Looking at your prior for cost, your cost attribute is very large (1744 and larger) and you are multiplying with a large prior (-0.33), this gives a huge utilities. Please use priors that make behavioural sense.