基于贝叶斯MCMC法的P-Ⅲ分布估参效率比较  

Research on the Efficiency of Parameter Estimation for P-Ⅲ Distribution Using Baysian MCMC Sampling Algorithm

作　　者：李扬 LI Yang

机构地区：[1]山西省水利水电科学研究院,山西太原030002

出　　处：《山西水利科技》2018年第4期4-8,共5页Shanxi Hydrotechnics

摘　　要：基于贝叶斯方法对P-Ⅲ分布的参数进行估计时,为解决复杂高维积分问题并降低计算量,采用MCMC方法得到参数样本。结合实例,对MH算法、DR算法和AM算法等3种常用MCMC方法的计算效率进行比较。结果表明,抽样算法的选择对MCMC方法的结果影响较大,而对频率曲线的拟合效果影响很小,与MH算法和DR算法相比,AM算法的收敛速度更快,计算效率更高。During the parameter estimating for P-Ⅲ distribution using Baysian theory,three kinds of MCMC sampling methods including MH algorithm,DR algorithm and AM algorithm were used to obtain the parameter samples of annual maximum flows,and their properties were compared in application.The results show that different sampling method has great influence on the performance of the MCMC method,and the frequency curve fitting effect has less relation to the type of sampling algorithm.Fast in convergence rate,high in computational efficiency and good effects in parameter estimation,AM algorithm was much better than MH and DR algorithm.

关 键 词：贝叶斯理论 MCMC法 MH算法 DR算法 AM算法 

分 类 号：P333.9[天文地球—水文科学]