One way to control for the heterogeneity in panel data is to allow for time-invariant, individual specific parameters. This fixed effect approach introduces many parameters into the model which causes the “incidental parameter problem”: the maximum likelihood estimator is in general...

In panel surveys, some observation units drop out before the end of the observation period. This panel attrition should not be ignored if it is related to the variables of interest. Hirano, Imbens, Ridder and Rubin propose the Additively Nonignorable (AN) attrition model to correct for the...