基于K-L变换的多维空间数据正态性检验方法及其应用

Normality Testing Method and Its Application of Multidimensional Spatial Data Based on K-L Transformation

分析了多维数据正态性检验方法及其适用性,研究了适用于多维空间数据正态性检验的基于K-L变换的检验法和基于最小生成树的检验法,并应用蒙特卡罗方法对两种方法进行了对比实验,结果证明基于K-L变换的检验法具有检验准确度高、鲁棒性强、运算速度快等优点。应用基于K-L变换的检验法对某区域连续6年的植被指数和降雨量的差异数据进行了多维正态性检验,进而分析了该区域6年中植被指数和降雨量变化的随机性。

The normality testing method of multidimensional data and its applicability were analyzed.The testing methods for the normality tests of multidimensional spatial data based on Karhunen-Loeve（K-L） transformation and on minimal spanning tree（MST） method respectively were studied and the contrast experiments of these methods were made by using Monte-Carlo method.The results showed that the method based on K-L transformation has advantages,such as high precision,strong robustness and fast calculation,etc.The multidimensional normality testing was made for the difference data of vegetable indices and rainfall in some region in continuous 6 years with the method based on K-L transformation and the randomness of vegetable indices and rainfall variation in 6 years in that region was also analyzed.