Abstract
In the analysis of count panel data regression, estimates of the maximum likelihood (ML) method for the Poisson fixed effects (PFE) model are influenced by outliers. So, the ML estimation is not suitable for dealing with the outliers in the count panel dataset. Therefore, this paper develops three robust estimation methods (M, S, and MM) for the PFE model to handle the outliers in count panel data. The Monte Carlo simulation study and two applications were performed for patents and COVID-19 to evaluate the proposed robust and non-robust traditional estimation methods of the PFE and negative binomial fixed effects (NBFE) models. Simulation and applications results indicated that non-robust estimates of the NBFE model are affected by outliers and therefore are not a good substitute for the PFE model when the dataset contains outliers. But the results of the robust estimates are better than PFE and NBFE estimates.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call