Spatial Autoregressive Model Estimation By Quasi-Maximum Likelihood And Transformation Methods For Panel Data Via Simulation

Abstract

This research deals with the study of the spatial autoregressive model for panel data, where the model parameters were estimated using two estimation methods (the Quasi maximum likelihood method, the transformation method) and with the presence of the spatial weight matrix modified according to the Rook adjacency criterion. The two methods were compared using the comparison criterion of the mean absolute percentage error (MAPE) with the aim of arriving at the best estimation method. Simulation experiments were also used on cross-sections (n=20) and for three different time values ​​(T=5,10,20) and for three different sample sizes (nt=100,200,400) and through the comparison criterion that leads us to the best estimation method, which is the transformation method (TTA).