Statistical Inference Spatial Econometric Model

The article aims to make statistical inferences of spatial econometric models, andmainly includes two aspects: on the one hand, under the econometric model of thespace in the symmetric weight matrix of zero diagonal elements, the existence anduniqueness of parameter for REMLE (restrictive max likelihood estimation) isdiscussed in the article; At the same time use the data simulation to verify thatREMLE is relative to the MLE (maximum likelihood estimation). on the other hand,the local influence analysis of the spatial econometric model for discussion, mainlydiscuss the mean-perturbation, so as to detect the strong influential observation andoutlier. The former apply the Bayesian method and the Orthogonal projection methodto construct the likelihood function. The structural key of the likelihood function isthe correction of the likelihood function. Then we apply the intermediate valuetheorem of continuous function to prove the existence and uniqueness and the prooffor two properties verify the feasibility of REMLE for the spatial econometric models.Compared with the traditional maximum likelihood estimation, The REMLE effect ismuch better than the effect for MLE, but there are also some limitations problem onthe REMLE. It is the method only applied to the symmetric weight matrix of zerodiagonal elements. The latter is mainly based on the parameters(,2), and use theFirst-Order method to construct the different local influential vector. Finally, usingthree simulated data, including perturnbation for neighboring points、perturnbation fornon-neighboring points and perturnbation for both neighboring and non-neighboringpoints, this paper simulation studies show that the local influence using the first-ordermethod can well handles the masking effect and drown effect. And the pros and consof the two methods make alongitudinal comparison. Thus, we can get that the localinfluential vector based on paramrter2can be more accurately reflect the stronglocal influential points and outliers. As an application, this paper analyzes a practicalexample to illustrate the usefulness of the proposed method.