How STATA handles missing data

[Robert Archer]

Good evening, I have a relatively straight forward question regarding STATA. I have been trying to find out what default procedure STATA is using to handle missing data in a negative binomial regression model? I see that a log likelihood iteration is repeated, which I feel suggests maximum likelihood, however I would prefer to know for sure. Thank you you in advance for any information that can be provided.

[Clyde Schechter]

Yes, the estimation of negative binomial regression models is done with maximum likelihood regression. But it is not full information maximum likelihood. In fact, observations with missing values for any variable in the model are omitted from the estimation sample, just as they are with all other Stata regression commands. So if you were thinking that because of maximum likelihood estimation you have dealt with the problem of missingness at random, I'm sorry to disappoint you, but no.

[Robert Archer]

Thank you much, this helps. The clarification of maximum likelihood and omitting variables is what I was looking for. Thanks again.

[Nick Cox]

It's a matter of omitting observations, not variables.