关键词: 半参数, 变系数, 广义空间滞后, 面板模型

Abstract: By constructing a semi-parametric varying-coefficient random effect panel model with spatial lag and spatial error, the flexibility and adaptability of the existing models are extended. It uses profile maximum likelihood method to construct both parametric and nonparametric estimators. The theoretical proof shows that all estimators have consistency and asymptotic normality under certain regular conditions. The numerical simulation shows that the estimators have good small sample properties, and the estimation accuracy increases with the increase of the sample capacity. The selection of the spatial weight matrix does not produce significant difference in the performance of the estimator, but under the condition of Case matrix, when the sample quantities are the same, the estimated deviation of spatial correlation coefficients increase with the increase of spatial weight structure complexity.

Key words: Semi-parametric, Varying-Coefficient, Generalized Spatial Lag, Panel Model