Incidental parameters problem in linear mixed models?

[Peter Schilling]

Dear Statalisters,

I currently fit a model of the following sort:

mixed dependent_variable covariates i.country_id i.industry_year_id || firm_id:

Given that the model is estimated using a maximum likelihood approach, I was wondering whether the incidental parameter problem arises? As I understand, the latter would be an issue in case I fitted a logit model for instance.

I appreciate your replies.

Best regards
Peter

[Fei Wang]

Peter, the incidental parameter problem arises when you estimate a model with many group-specific parameters (like individual fixed effects) using ML approach, in particular when the sample size within each group is relatively fixed as the number of groups grows. In my field (economics), the problem almost only appears in non-linear (panel data) model with fixed effects, because the fixed effects parameters can be easily eliminated in linear setting (like first-differencing). I'm not familiar with linear mixed model, but I assume the incidental parameter problem won't be a big issue if the group-specific parameters can be wiped out in advance. (Or if the linear model is estimated using least squares method, it won't be a problem either.)