1
$\begingroup$

I have a dataset of around 30,000 people where each chooses one of 4 items: A, B, C or D. People are nested within the 600 areas. I want to fit a multinomial logit model, where random intercepts are calculated for each of the 600 areas for each option. i.e. we allow that people in some areas may have a greater tendency to pick A. I also want to include area level predictors and possibly individual level predictors.

What software would be suitable for fitting such a model? The mlogit package in R does not appear to allow alternative specific random intercepts and various other packages I've tried in Stata and R do not seem to have exactly this set of options. I'm open to trying out any software if it will do the job.

$\endgroup$
1
  • $\begingroup$ Stan has a steep learning curve, but you can fit a pretty diverse set of models with it, including this one. $\endgroup$ – Sycorax Jan 5 '15 at 16:16
1
$\begingroup$

As mentioned in the comments there is Stan, BUGS, and JAGS. Any of these will allow you to fit the model you are describing. Here is a link of example code in BUGS and Stan: https://github.com/stan-dev/example-models/wiki/BUGS-Examples-Sorted-Alphabetically

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.