空间权重矩阵的生成方法分析与实验

Analysis and Application on the Specification Methods of the Spatial Weight Matrix

区域空间动力学过程 (例如疾病传播、区域社会经济可持续发展等 )所涉及的数据样本和状态变量都是以区域多边形形式存在的。区域群发展在空间上的此起彼伏和相互影响则是通过区域之间相互联系得以实现的 ,空间权重矩阵用以传载这一作用过程。因此 ,构建区域空间连接权重矩阵是建立区域空间动力学方程的核心问题之一。本文分析了空间权重矩阵的不同生成方法 ,并对空间权重矩阵的一种形式 。

The study of regional spatial weight matrix is of very importance and indispensability in the research of regional spatio temporal process in spatial econometics.Nowadays,when dealing with the regional data, we turn our attentions from pure information storing or staticstical analysis to exploring their growing and developing law by spatio temporal dynamics models. In time dimension, they have their own movement law, but in space the effects from their adjacent regions, that is, spatial dependence, can't be disregarded ,either. And this spatial dependence is expressed by the spatial weight matrix in the spatio temporal process model. Thus constructing the regional spatial weight matrix is one of the key steps of creating regional spatial dynamics equations. In this paper we have reviewed a series of specification forms of the spatial weight matrix, such as binary contiguity matrix, distance weight matrix, a combination of distance and perimeter weight matrix, a combination of distance and area weight matrix, etc. Therefore spatial researchers are at a loss when facing the weight matrix choosing problem and resort to the simplest binary contiguity matrix. But the simple choice of spatial contiguity matrix is very disputable. Hence, in practical application how to choose the form of spatial weight matrix calls for further studies. Although the use of binary contiguity matrix in constructing a sptial process model is dubitable, its simpleness helps the analyst to construct the sptial process model and analyse the data easily at first stage. As for a map of reams of analysis regions, it's fussy and likely to make mistakes to construct a binary contiguity matrix manually. In the paper we come up with the detailed algorithm by GIS software automatically to generate the binary contiguity matrix, which has been validated to be feasible and robust.