Dear Statalist users,


I am analysing the probability of doctors to accept a clinical job offer using a panel data with 3,600 observation over 13 time periods (Data were collected from a discrete choice model). The dependent variable is coded 0 or 1, with 1 when the event occurs (i.e. doctors accept the job offer).

The total number of 1 is 210 (5.83%), whilst 0 represents 3,390 observations (94.17%), which seems to correspond to rare events data.

I used a random-effect Probit model but to no avail as the model did not converge, and two of my independent variables predicted the outcome perfectly. I used firthlogit, however this does not account for the panel nature of the data. Is there an alternative to "firthlogit" for panel data with rare events?

Your help would be much appreciated.Thanks
Zoé