Statistical analysis of zero-inflated Rayleigh distribution
-
摘要: 现实生活中,许多领域的数据集存在零值显著偏多的现象,传统分布模型难以有效拟合此类数据。针对这一问题,构建零膨胀瑞利分布模型,详细推导该模型的统计性质,包括极大似然估计、客观先验及后验分析,并设计相应的抽样算法,以更好地描述具有零膨胀特征的复杂数据。通过数值仿真实验,对不同样本容量和参数真值下模型参数进行估计和验证。实验结果表明,提出的抽样算法能够有效估计零膨胀瑞利分布中的参数,为实际应用提供了理论支持和计算工具。研究拓展了零膨胀模型的理论框架,为众多领域的数据分析提供新的解决方案。
-
关键词:
- 零膨胀瑞利分布 /
- 极大似然估计 /
- 客观贝叶斯分析 /
- reference先验
Abstract: In real-world scenarios, datasets in many fields exhibit a significant excess of zero values, which traditional distribution models struggle to fit effectively. To address this issue, a zero-inflated Rayleigh distribution model was constructed. The statistical properties of the model, including maximum likelihood estimation, objective prior and posterior analysis, were derived in detail, and a corresponding sampling algorithm was designed to better describe complex data with zero-inflation characteristics. Through numerical simulation experiments, the model parameters were estimated and validated using different sample sizes and true parameter values. The experimental results demonstrate that the proposed sampling algorithm can effectively estimate the parameters of the zero-inflated Rayleigh distribution, providing theoretical support and computational tools for practical applications. The study not only expands the theoretical framework of zero-inflated models but also offers new solutions for data analysis in numerous fields. -
表 1 参数估计量的均值
Table 1. Mean of parameter estimation
$ q $ $ \eta $ $ n $ $ {q}_{{\mathrm{MLE}}} $ $ {\eta }_{{\mathrm{MLE}}} $ $ {q}_{{\mathrm{J}}} $ $ {\eta }_{{\mathrm{J}}} $ $ {q}_{{\mathrm{R}}1} $ $ {\eta }_{{\mathrm{R}}1} $ $ {q}_{{\mathrm{R}}2} $ $ {\eta }_{{\mathrm{R}}2} $ 0.3 1 50 0.300 1 1.022 7 0.315 8 1.029 6 0.293 6 1.028 1 0.317 9 1.053 4 100 0.295 6 0.943 2 0.295 8 0.936 2 0.297 9 0.943 0 0.303 4 0.955 0 200 0.303 2 0.968 3 0.302 7 0.965 7 0.302 2 0.968 5 0.298 4 0.974 6 3 50 0.303 5 3.059 2 0.307 6 3.081 5 0.302 9 3.067 5 0.303 7 3.186 6 100 0.297 0 2.800 8 0.304 4 2.823 9 0.301 4 2.844 5 0.300 6 2.876 7 200 0.302 3 2.899 8 0.301 8 2.907 3 0.298 9 2.902 9 0.303 9 2.922 6 0.7 1 50 0.682 0 1.019 5 0.609 6 1.096 0 0.694 7 1.028 3 0.690 1 1.108 2 100 0.688 1 0.945 4 0.664 6 0.967 6 0.693 5 0.931 3 0.702 8 0.973 1 200 0.693 6 0.968 9 0.656 0 0.992 1 0.696 3 0.973 6 0.694 0 0.988 9 3 50 0.694 7 3.045 1 0.699 0 3.058 8 0.708 9 3.030 9 0.692 9 3.284 9 100 0.695 2 2.794 9 0.696 7 2.813 9 0.688 6 2.820 2 0.698 2 2.910 1 200 0.693 4 2.885 9 0.698 0 2.885 1 0.698 0 2.911 6 0.698 0 2.999 1 表 2 参数估计量的均方误差
Table 2. Mean square error of parameter estimation
$ q $ $ \eta $ $ n $ $ {q}_{{\mathrm{MLE}}} $ $ {\eta }_{{\mathrm{MLE}}} $ $ {q}_{{\mathrm{J}}} $ $ {\eta }_{{\mathrm{J}}} $ $ {q}_{{\mathrm{R}}1} $ $ {\eta }_{{\mathrm{R}}1} $ $ {q}_{{\mathrm{R}}2} $ $ {\eta }_{{\mathrm{R}}2} $ 0.3 1 50 0.003 9 0.002 4 0.008 1 0.002 9 0.003 4 0.002 5 0.005 3 0.005 2 100 0.001 8 0.004 1 0.002 1 0.004 7 0.002 4 0.004 0 0.002 2 0.003 0 200 0.001 1 0.001 6 0.001 6 0.001 6 0.000 9 0.001 4 0.001 0 0.001 1 3 50 0.004 2 0.024 1 0.003 6 0.027 5 0.003 6 0.020 7 0.004 4 0.052 1 100 0.002 4 0.047 3 0.002 0 0.037 8 0.001 7 0.030 1 0.002 0 0.022 4 200 0.000 7 0.013 9 0.001 1 0.012 2 0.000 9 0.013 4 0.000 8 0.010 5 0.7 1 50 0.004 0 0.008 9 0.017 0 0.050 3 0.004 8 0.011 3 0.005 1 0.023 3 100 0.002 4 0.007 2 0.008 2 0.012 4 0.002 1 0.008 7 0.002 6 0.005 3 200 0.001 1 0.003 2 0.008 8 0.009 7 0.001 0 0.002 9 0.001 1 0.002 3 3 50 0.004 3 0.095 0 0.003 6 0.095 7 0.003 6 0.098 0 0.004 3 0.186 5 100 0.002 1 0.078 1 0.002 9 0.091 9 0.002 6 0.063 2 0.001 8 0.052 0 200 0.001 1 0.034 2 0.001 0 0.036 8 0.001 1 0.030 4 0.001 1 0.020 9 -
[1] 肖翔, 吴懿祺, 古晞. 基于指数分布混合治愈模型的客观贝叶斯分析[J] . 上海工程技术大学学报, 2021, 35(4): 395 − 400. doi: 10.3969/j.issn.1009-444X.2021.04.015 [2] 徐昕. 零膨胀广义泊松模型的推广及其在精算中应用[J] . 数理统计与管理, 2020, 39(6): 1010 − 1021. doi: 10.13860/j.cnki.sltj.20200930-007 [3] 张良超, 周金亮, 温利民. 零膨胀泊松模型中风险参数的贝叶斯估计[J] . 江西师范大学学报(自然科学版), 2020, 44(3): 269 − 274. doi: 10.16357/j.cnki.issn1000-5862.2020.03.10 [4] Lambert D. Zero-inflated Poisson regression, with an application to defects in manufacturing[J] . Technometrics, 1992, 34(1): 1 − 14. doi: 10.2307/1269547 [5] Greene W H. Accounting for excess zeros and sample selection in Poisson and negative binomial regression models[D] . New York: New York University, 1994. [6] 程平平. 零膨胀指数分布的客观贝叶斯分析[D] . 上海: 华东师范大学, 2013. [7] 盛建为, 钱夕元. 零膨胀对数级数分布的参数估计[J] . 华东理工大学学报(自然科学版), 2019, 45(3): 507 − 510. doi: 10.14135/j.cnki.1006-3080.20180419003 [8] De Freitas Costa E, Schneider S, Carlotto G B, et al. Zero-inflated-censored Weibull and gamma regression models to estimate wild boar population dispersal distance[J] . Japanese Journal of Statistics and Data Science, 2021, 4(2): 1133 − 1155. doi: 10.1007/s42081-021-00124-0 [9] Bodhisuwan R, Kehler A. The zero-inflated negative binomial-exponential distribution and its application[J] . Lobachevskii Journal of Mathematics, 2021, 42(2): 300 − 307. doi: 10.1134/S1995080221020062 [10] 刘梦瑶, 肖翔. 零一膨胀几何分布的统计分析及应用[J] . 上海工程技术大学学报, 2024, 38(2): 196 − 204. doi: 10.3969/j.issn.1009-444X.2024.02.013 [11] Pishgar-Komleh S H, Keyhani A, Sefeedpari P. Wind speed and power density analysis based on Weibull and Rayleigh distributions (a case study: firouzkooh county of Iran)[J] . Renewable and Sustainable Energy Reviews, 2015, 42: 313 − 322. doi: 10.1016/j.rser.2014.10.028 [12] 郎献军, 何玉辉, 唐进元, 等. 基于磨粒突出高度为瑞利分布的磨削力模型[J] . 中南大学学报(自然科学版), 2014, 45(10): 3386 − 3391. [13] Sarti A, Corsi C, Mazzini E, et al. Maximum likelihood segmentation of ultrasound images with Rayleigh distribution[J] . IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, 2005, 52(6): 947 − 960. doi: 10.1109/TUFFC.2005.1504017 [14] 茆诗松, 汤银才. 贝叶斯统计[M] . 2版. 北京: 中国统计出版社, 2012: 92−97. -
25-0025FL.pdf
-
下载: