Maximum Simulated Likelihood Estimation with Correlated Observations: A Comparison of Simulation Techniques Used for Spatial Econometric Models

摘要: Spatial econometric model is a powerful tool for analyzing regional issues, particularly those involving land use, travel, and demographics. Complex spatial models are normally intractable require special estimation methods. Maximum simulated likelihood (MSLE) one of the methods that become popular in recent years. MSLE routines new software releases often provide several optional simulation techniques. It important analysts understand relative performance different techniques under data circumstances, especially studies, where observations spatially correlated. This paper studies with correlated observations. found quasi Monte-Carlo (QMC) impose strong periodic correlation pattern across While some forms sequencing, such as scrambled Halton, Sobol Faure, can sever correlations dimensions error-term integration, they cannot remove carries through When set’s true patterns clearly differ from patterns, may inefficient finite samples, statistical identification parameters suffer. Fortunately, it most cases even when correlated, QMCs hybrid be preferred to pseudo method, thanks their better coverage. The findings reported here offer an supplement existing should prove valuable future work requires