Title: Adjusted maximum likelihood inference for spatial panel data models
Authors: Federico Martellosio - University of Surrey (United Kingdom) [presenting]
Abstract: In a likelihood framework, an often successful way to deal with incidental parameters is to ``adjust'' the profile score of the parameter of interest. We consider the adjusted quasi-maximum likelihood estimator (QMLE) of the spatial parameter in a spatial panel model with individual and/or time fixed effects. The adjusted QMLE coincides with the QMLE if covariates are not present in the model. When covariates are present, the adjusted QMLE can be more accurate than the QMLE. Saddlepoint confidence intervals for the spatial autoregressive parameter based on the adjusted QMLE are proposed. In simulation, they perform very well against other higher-order methods.