提出了不同于计算相关系数的列序分析的方法,思路来源于变量序列间的几何形状接近程度,几何形状接近的序列,度量其靠近程度的指数应大,反之就小.将靠近程度大小按序排列,就得到变量序列间的联系程度的名次.从实用出发,针对不同性质的数据,提出了4种列序度计算方案.选取了计算加权平方距离的方案对大气污染及其有关的气象要素数据进行了实例计算.通过计算列序度并与相关系数对照,可以更可靠地来使用某个气象要素制作预报,另外,从数学上证明列序分析与绝对值关联度、欧氏距离间的关系,由此论证了列序分析的数学根基.数学推导和数值计算都证实了列序分析的可用性,尤其在量测为小样本的情况下更有使用价值.%The method for calculating time sequences through ordinal analysis,unlike correlation coefficient calculation,is presented,in which the idea comes from the closeness of the geometry between the variable sequences.For the sequences with the geometry closer to each other,the index that measures the closeness is bigger and vice versa.Putting the closeness in numerical order,the ranking of degree of contact between the variable sequences can be obtained.Four sequence calculation methods for different nature of the data from the practical point of view are proposed.Choosing the scheme of calculating the weighted squared distance,the related calculation is conducted with air pollution and its related meteorological elements data.By calculating the closeness degree of sequences and comparing with the coefficient method,it is concluded that by the method it is more reliable to use a meteorological element in forecasting.The relationship between the ordinal analysis and the correlation degree of absolute values,as well as the Euclidean distance mathematically,is proved,and the mathematical foundation of ordinal analysis is thus demonstrated.Mathematical derivation and numerical calculations both confirm the availability of ordinal analysis; especially it is useful in the case of measurements for small samples.
展开▼
