基于非对称复合矩阵的房价空间自相关性研究

A STUDY ON THE SPATIAL AUTOCORRELATION IN HOUSING PRICES BASED ON ASYMMETRIC COMPOUND MATRIX

分析地理空间权重矩阵在捕捉房价空间自相关性时的局限性,基于高斯核函数针对截面数据提出非对称复合空间权重矩阵,使用这两种矩阵对比分析我国城际住宅价格的空间分布格局。结果显示非对称复合空间权重矩阵比地理空间权重矩阵揭示的住宅价格全局空间正相关性显著,住宅价格总体上存在空间集聚格局,集聚程度呈递减趋势;大部分城市住宅价格的局部空间同质性显著,"高—高"型显著性集聚位于东部沿海城市且范围变化不大,"低—低"型显著性集聚由中西部城市扩大至东北部城市,少数城市存在局部空间异质性格局但不显著,住宅价格呈现局部空间二元结构,并解析住宅价格空间分布格局的成因。

This article analyzes the limitations of geographical weight matrix on spatial autocorrelation in housing prices and proposes an asymmetric compound weight matrix for cross-sectional data on the basis of Gaussian kernel function. These two kinds of matrix are used to compare and analyze the spatial distribution pattern of intercity housing prices in China. The results show that the overall spatial positive correlation of the asymmetric compound weight matrix is more significant than that of the spatial weight matrix. In general, there is a spatial agglomeration pattern in housing prices, and the agglomeration degree is de-creasing. Residential prices of most cities has significant local spatial homogeneity. The cities of 野 H -H冶（agglomerative type） are located in the eastern coastal areas and change slightly, while the cities of 野 L -L冶 expand from the mid-west to the north-east. Residential prices of few cities has local spatial heterogeneity which is not significant. Local spatial distribution in resi-dential prices presents binary structures, and the reasons resulting in this spatial distribution are discussed.