On Gmm Is Estimated That In The Mrsar Model Overidentified Linear Models

Generalized method of moments (GMM), an important estimation method, is commonly used in parameter estimation of economic and statistical models. The most two popular spatial econometric models are mixed regressive-spatial autoregressive (MRSAR) model and over-identifying linear model. In this paper, we study the GMM estimators of both the models refereed to above, and find their good characters, which based on the advantage of GMM.This paper is organized as follows.The first chapter is an introduction to the history of GMM theory and some basic notions used in the paper.The second chapter discusses the GMM of MRSAR model. In the MRSAR model, GMM can be computationally simpler and its efficiency property is derived. And it shows that when the disturbances are distributed by linear and quadratic conditions, the best generalized method of moments (BGMME) exists. Then empirical counterparts is deduced, we call it FGMME, and show that FGMME has the same limiting distribution as the corresponding BGMME.The last chapter considers the over-identifying linear model. It studies two versions of the GMM over-identifying restrictions tests and the concept of approximate slopes is employed to compare their power properties globally. Without the restrictions of the martingale difference assumption, it is found that the GMM over-identifying test with the v-class auto-covariance matrix is more powerful than the test using the v-class non-centering auto-covariance matrix one. It relaxes conditions in essay [1], and promotes its conclusions.