Page 1 of 1

eStat Error running 1levelundordered multinomial logit

Posted: Wed Oct 28, 2020 12:14 pm
by kofidlehman
Hello,
I get an error running the template for the 1levelunorderd multinomial logit with eStat. I have attached the errors reported in Stat-JR and the commands reported in the command window. I would appreciate any help available figuring out what is going on. Thanks.
Kofi

Re: eStat Error running 1levelundordered multinomial logit

Posted: Thu Oct 29, 2020 7:21 pm
by ChrisCharlton
Looking at the model and subsequent error messages it looks like it is missing the prior for alpha_2:

Code: Select all

model
{
    for(i in 1:length(couple_ideal)) {
       dummy[i] ~ dcatun3(p1[i], p2[i], p3[i],r1[i],r2[i],r3[i])

       ln(p1[i]) <- constant[i] * alpha_0
       ln(p2[i]) <- constant[i] * alpha_1
       ln(p3[i]) <- constant[i] * alpha_2
    }
    # Priors
    alpha_0 ~ dflat()
    alpha_1 ~ dflat()
}
I will need to go through the template to try to work out why this is.

Re: eStat Error running 1levelundordered multinomial logit

Posted: Thu Oct 29, 2020 7:35 pm
by ChrisCharlton
Can you please let me know how many categories you have in your unordered response, and what you have chosen as the base category? I have a feeling that it isn't correctly excluding the base from the dcatun3 line, but is for the priors.

Re: eStat Error running 1levelundordered multinomial logit

Posted: Fri Oct 30, 2020 12:43 pm
by kofidlehman
Hi Chris,
Thank you for the prompt response. I have three categories - 1, 2 and 4. 4 is the excluded category.

Re: eStat Error running 1levelundordered multinomial logit

Posted: Fri Oct 30, 2020 7:55 pm
by kofidlehman
Hi Chris,
Following your comment, I recoded the categories of the dependent variable to be 1,2 and 3 with 3 as the excluded category and that appears to fix the problem.
Kofi

Re: eStat Error running 1levelundordered multinomial logit

Posted: Tue May 09, 2023 8:34 am
by celinedion
Himapquest driving directions
To fix this, you can add a prior distribution for alpha_2 similar to how priors are assigned for alpha_0 and alpha_1. For example, you can use a flat prior as follows:
alpha_2 ~ dflat()