P—Ⅲ型参数估计方法的研究

本文应用Moute Carlo试验,对P-Ⅲ型分布的参数估计方法作了详细地比较,包括矩法、极大似然法、适线法、转换法和概率权重法。在矩法中又分析对比了六种不同的C_s修正公式。同时也考虑了四种不同的转换法,并着重讨论经验频率公式的选用。通过大量的计算工作,作者建议采用极大似然法(当C_s<2.0时)或概率权重法(当C_s>2.0时)来估算P-Ⅲ型的参数和设计洪水,用Bobee和Robitaille公式来计算C_s的矩法在实用上也是可行的。它们都要比适线法好。Weibull公式是有偏的,不宜继续使用,作者提议应该对我国现有的设计洪水计算规范作出重新的论证和评价。

Fitting the Pearson Type 3 Distribution

Based upon the Monte Carlo experiments, the different parameter estimation methods are discussed and compared in the case of Pearson Type 3 distribution, including methods of moments, maximum likelihood, curve fitting, probability weighted moments and transformation methods. For the method of moments, six different bias correction factor of skewness are considered. Four various transformation methods and five different plotting position fromulae are also discussed. Through a large number of computation work, it is suggested to use maximum likelihood (when Cs<2.0) or probability weighted moments (when Cs≥2.0) to estimate parameters and quantiles. The method of moments, with skewness corrected by Bobee and Robitaille's factor is a practical alternative. All of them are superior to the graphical curve fitting methods. The weibull plotting position formulae are biased and have large variance. It may be suggested that the flood design guidelines currently in use in China are in need of reevaluation.