multilevel data and rare events

[Caroline Wilson]

Hi everyone,

I have data with a nested structure, and am running two separate multilevel logistic regression models (my dependent variable is binary). The problem is that, for both models, the number of values of the ‘1’ category in the dependent variable is low. In the first model, 62 of 5,807 (1%) of the observations had a value of 1; the rest had a value of 0. For the second model, 41 of 2,347 observations (1.7%) had the value of 1; the rest had the value of 0.

I am aware of models for rare events data such as PMLE (firthlogit in Stata), but I don’t believe they work when the dataset has a nested structure. Is there a way to estimate such a model in Stata? If not, what is best – still fit the multilevel model or fit a PMLE model ignoring the clustering?

Or is there a better way?

Thank you in advance for any comments,
Caroline

[Caroline Wilson]

I have linked this question to a previous forum I started which has more description of the dataset (here: http://www.statalist.org/forums/foru...-effects-model)
I apologize for the multiple postings.

[Oded Mcdossi]

I don't know what is the exact formula behind firthlogit and exlogistic and perhaps some of the econometricians in the forum would help to identify that (the author of firthlogit Joseph Coveney is an active participant in this forum and perhaps he can help with that). I guess having these details you can use the meglm command to estimate this model under the nesting assumption. But this is just general thoughts as I haven't used these models before.