Volume 6, Issue 3 (Vol. 6, No. 3 2020)
mmr 2020, 6(3): 405-416 |
Back to browse issues page
Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
Sharafi M, Tavangar F, Enami S. Three-Parameter Kappa Distribution and its Fitting to the Whole Monthly Rainfall Data of Abali Station in Tehran Province. mmr 2020; 6 (3) :405-416
URL: http://mmr.khu.ac.ir/article-1-2846-en.html
URL: http://mmr.khu.ac.ir/article-1-2846-en.html
1- Shiraz University , msharafi@shirazu.ac.ir
Abstract: (2398 Views)
Introduction
The kappa distribution was first introduced by Mielke (1973) and Mielke and Johnson (1973) for describing and analyzing precipitation data. This distribution is positively skewed and is widely applied when studying precipitation, wind speed and the stream flow data in hydrology. The kappa distribution has some advantages over gamma and log-normal distributions in fitting historical rainfall. Data It is because, unlike the latter two distributions, it has closed forms for the cumulative distribution function and quantile function. Due to this important feature, the kappa distribution attracts the attention of several researchers. Park et al. (2009) introduced the three-parameter kappa distribution and provided a description of the mathematical properties of the distribution and estimated the parameters by three methods. Also, they illustrated its applicability for rainfall data from Seoul, Korea. In this paper, we study the distribution and the estimation methods for the parameters considered by Park et al. (2009), and propose a new estimation method. Then, we will compare these estimation methods using a Monte Calro simulation study and a real dataset.
Material and methods
In this scheme, first we consider the three-parameter kappa distribution and study some of its properties and then estimate the parameters of the distribution by four methods. These methods are method of moment (MM), L-moments (LM), maximum likelihood (ML) and maximum product of spacing method (MPS). Using a Monte Carlo simulation study and a real data set, performance of these methods are compared.
Results and discussion
Comparing the performance of the proposed estimation methods in terms of bias and root of mean squares error (rmse), it can be concluded that the MPS method has a better performance due to its lower bias and rmse. The Kolmogorov-Smirnov test is applied for goodness-of-fit test in the three-parameter kappa distribution to the whole monthly rainfall data of Abali station in Tehran province. The results demonstrate that the MPSE method leads to better results than other mentioned methods.
Conclusion
The following conclusions were drawn from this research.
The kappa distribution was first introduced by Mielke (1973) and Mielke and Johnson (1973) for describing and analyzing precipitation data. This distribution is positively skewed and is widely applied when studying precipitation, wind speed and the stream flow data in hydrology. The kappa distribution has some advantages over gamma and log-normal distributions in fitting historical rainfall. Data It is because, unlike the latter two distributions, it has closed forms for the cumulative distribution function and quantile function. Due to this important feature, the kappa distribution attracts the attention of several researchers. Park et al. (2009) introduced the three-parameter kappa distribution and provided a description of the mathematical properties of the distribution and estimated the parameters by three methods. Also, they illustrated its applicability for rainfall data from Seoul, Korea. In this paper, we study the distribution and the estimation methods for the parameters considered by Park et al. (2009), and propose a new estimation method. Then, we will compare these estimation methods using a Monte Calro simulation study and a real dataset.
Material and methods
In this scheme, first we consider the three-parameter kappa distribution and study some of its properties and then estimate the parameters of the distribution by four methods. These methods are method of moment (MM), L-moments (LM), maximum likelihood (ML) and maximum product of spacing method (MPS). Using a Monte Carlo simulation study and a real data set, performance of these methods are compared.
Results and discussion
Comparing the performance of the proposed estimation methods in terms of bias and root of mean squares error (rmse), it can be concluded that the MPS method has a better performance due to its lower bias and rmse. The Kolmogorov-Smirnov test is applied for goodness-of-fit test in the three-parameter kappa distribution to the whole monthly rainfall data of Abali station in Tehran province. The results demonstrate that the MPSE method leads to better results than other mentioned methods.
Conclusion
The following conclusions were drawn from this research.
- The Monte Carlo simulation shows that the maximum product spacing method, which is proposed in this paper, is the best method for estimating the parameters of the three-parameter kappa distribution.
- The statistics and p-value of the Kolmogorov–Smirnov test show that the three-kappa distribution with the MPS method of estimation has better fit than the other methods.
Keywords: Three-parameter kappa distribution, Maximum likelihood estimator, L-Moments estimator, Maximum product of spacings estimator.
Type of Study: S |
Subject:
stat
Received: 2018/09/26 | Revised: 2020/12/14 | Accepted: 2019/01/14 | Published: 2020/11/30 | ePublished: 2020/11/30
Received: 2018/09/26 | Revised: 2020/12/14 | Accepted: 2019/01/14 | Published: 2020/11/30 | ePublished: 2020/11/30
Send email to the article author
| Rights and permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |
