水文分析计算中两种正态变换方法的比较研究

Comparisons of Two Normalization Approaches in Hydrologic Analytic Calculation

采用Monte Carlo统计试验方法对水文分析中常用的Box Cox变换和多项式正态变换的统计性能进行了对比分析。以P Ⅲ分布样本作为原始样本,采用参数检验和拟合分布检验方法比较了两种方法变换后样本的正态性,分析了反变换后对水文设计值的估计效果。研究表明,Box Cox变换后样本的正态性更优,若以反变换后对水文设计值估计的统计性能评价,则以多项式正态变换的效果更优。

Comparisons of statistical features between two normalization approaches, namely Box-Cox and Polynomial normality transforms (PNT), were presented by means of Monte-Carlo technique. The Pearson-type-three probability distribution (P-Ⅲ) was taken as the population distribution from which the original samples were generated. The normality of the transformed samples by the two transforming approaches was investigated through parametric and goodness-of-fit test. The statistical properties of the estimations on quantites for the population distribution or original data were studied and compared by reverse transforming. Results indicate that the Box-Cox approach provides better normality for data transforms, but in the case of evaluation by the statistical properties of the estimations on quantiles through reverse transform, the PNT is superior to Box-Cox method.