The article discusses how to estimate panel data models with fixed effects. It shows that some commonly used estimators are not consistent under certain conditions. The researchers compare different estimators and find that the FEML estimator is the most efficient. They also compare different approaches for estimating panel data models and covariance parameters.