Hi everyone
I need to run a mixed effects multinomial logistic regression with random intercepts at the site of recruitment (15 sites). My dependent variable is categorical with 3 unordered levels. I was wondering what would be the best option to estimate the effect of the site of recruitment on the outcome.
With linear models in the past I estimated the variance partition coefficient (VPC) to state how much of the variance could be attributable to differences within the groups.
I thought that one possibility might be exponentiating the variance of the random effects but that would just inform about the reference level of the site (let's say Amsterdam), isn't that correct?
I used xtmlogit. The code follows.
xtset site
xtmlogit pred_LC2 i.sex c.age ib6.ethn_allsites, base(2) vce(cluster site)
Thanks.
I need to run a mixed effects multinomial logistic regression with random intercepts at the site of recruitment (15 sites). My dependent variable is categorical with 3 unordered levels. I was wondering what would be the best option to estimate the effect of the site of recruitment on the outcome.
With linear models in the past I estimated the variance partition coefficient (VPC) to state how much of the variance could be attributable to differences within the groups.
I thought that one possibility might be exponentiating the variance of the random effects but that would just inform about the reference level of the site (let's say Amsterdam), isn't that correct?
I used xtmlogit. The code follows.
xtset site
xtmlogit pred_LC2 i.sex c.age ib6.ethn_allsites, base(2) vce(cluster site)
Thanks.
