水文序列变异点检验方法的性能比较分析  

Comparison and Analysis on the Performance of Hydrological Time Series Change-point Testing Methods

作　　者：雷红富[1] 谢平[1] 陈广才[1] 李晶[1] 

机构地区：[1]武汉大学水资源与水电工程科学国家重点实验室,湖北武汉430072

出　　处：《水电能源科学》2007年第4期36-40,共5页Water Resources and Power

基　　金：国家自然科学基金资助项目(50579052)

摘　　要：通过设计的水文变异序列生成器生成满足P-型分布及统计参数,用统计实验的方法比较分析了10种常用水文序列变异点检验方法对3类变异序列的性能检验。结果表明,秩和检验、Brown-Forsythe、有序聚类、滑动T检验、Bayesian、里海哈林检验方法对均值变异类序列检验效果明显,滑动F检验法对变差系数CV变异序列检验效率优于其他方法,各方法对偏态系数CS变异序列检验效率较低。In order to compare and analyze the performance of hydrological time variation series change-point testing method, three types of hydrological series are generated by hydrological variation series generator, and statistical experiment is used to analyze the performance of the 10 variation series testing methods. The results indicate that Rank Sum test,Brown-Forsythe test,Order Cluster analysis,sliding T test, Bayesian method, and Lee-Heghinian method can test mean variation series exactly to a certain extent, and sliding F Test have better ability to test Cvvariation series,but every method is not sensitive to the Cs variation series.

关 键 词：水文序列 变异点 检验方法 性能 统计实验 

分 类 号：P333[天文地球—水文科学]