Kybernetika 58 no. 6, 883-902, 2022

Improved inference for the generalized Pareto distribution under linear, power and exponential normalization

Osama Mohareb Khaled, Haroon Mohamed Barakat and Nourhan Khalil RakhaDOI: 10.14736/kyb-2022-6-0883

Abstract:

We discuss three estimation methods: the method of moments, probability weighted moments, and L-moments for the scale parameter and the extreme value index in the generalized Pareto distribution under linear normalization. Moreover, we adapt these methods to use for the generalized Pareto distribution under power and exponential normalizations. A simulation study is conducted to compare the three methods on the three models and determine which is the best, which turned out to be the probability weighted moments. A new computational technique for improving fitting quality is proposed and tested on two real-world data sets using the probability weighted moments. We looked back at various maximal data sets that had previously been addressed in the literature and for which the generalized extreme value distribution under linear normalization had failed to adequately explain them. We use the suggested procedure to find good fits.

Keywords:

probability weighted moments, generalized Pareto distribution, generalized extreme value distribution, method of moments, L-moments, linear-power-exponential normalization

Classification:

62F10, 62F03

paper.pdf

References:

  1. A. A. Balkema and L. de Haan: Residual life time at great age. Ann. Prob. 2 (1974), 5, 792-804.   CrossRef
  2. H. M. Barakat, E. H. Nigm and E. O. Abo Zaid: Asymptotic distributions of record values under exponential normalization. Bull. Belg. Math. Soc. Simon Stevin 26 (2019), 743-758.   DOI:10.36045/bbms/1579402820
  3. H. M. Barakat, E. M- Nigm and H. A. Alaswed: The Hill estimators under power normalization. App. Math. Modell. 45 (2017), 813-822.   DOI:10.1016/j.apm.2017.01.028
  4. H. M. Barakat, E. M. Nigm and M. E. El-Adll: Comparison between the rates of convergence of extremes under linear and under power normalization. Stat. Pap. 51 (2010), 1, 149-164.   DOI:10.1007/s00362-008-0128-1
  5. H. M. Barakat, E. M. Nigm and O. M. Khaled: Extreme value modeling under power normalization. App. Math. Modell. 37 (2013), 10162-10169.   DOI:10.1016/j.apm.2013.05.045
  6. H. M. Barakat, E. M. Nigm and O. M. Khaled: Statistical modeling of extremes under linear and power normalizations with applications to air pollutions. Kuwait J. Sci. 41 (2014), 1, 1-19.   CrossRef
  7. H. M. Barakat, E. M. Nigm and O. M. Khaled: Bootstrap method for central and intermediate order statistics under power normalization. Kybernetika 51 (2015), 923-932.   DOI:10.14736/kyb-2015-6-0923
  8. H. M. Barakat, O. M. Khaled and N. Khalil Rakha: Modeling of extreme values via exponential normalization compared with linear and power normalization. Symmetry 12 (2020), 11, 1876.   DOI:10.3390/sym12111876
  9. H. M. Barakat, E. M. Nigm, O. M. Khaled and H. A. Alaswed: The counterparts of Hill estimators under power normalization. Special Issue J. App. Stat. Sci. (JASS) 22 (2016), 1-2, 87-98.   CrossRef
  10. H. M. Barakat, A. R. Omar and O. M. Khaled: A new flexible extreme value model for modeling the extreme value data, with an application to environmental data. Stat. Probab. Lett. 130 (2017), 25-31.   DOI:10.1016/j.spl.2017.07.002
  11. H. M. Barakat, E. M. Nigm, O. M. Khaled and H. A. Alaswed: The estimations under power normalization for the tail index, with comparison. AStA Adv. Stat. Anal. 102 (2018), 3, 431-454.   DOI:10.1007/s10182-017-0314-3
  12. H. M. Barakat, E. M. Nigm, O. M. Khaled and F. M. Khan: Bootstrap order statistics and modeling study of the air pollution. Comm. Statis.-Sim. Comp. 44 (2015), 1477-1491.   CrossRef
  13. S. BuHamra, N. Al-Kandari and M. Al-Harbi: Parametric and nonparametric bootstrap: An analysis of indoor air data from Kuwait. Kuwait J. Sci. 45 (2018), 2, 22-29.   CrossRef
  14. E. Castillo, A. S. Hadi, N. Balakrishnan and J. M. Sarabia: Extreme Value and Related Models with Applications in Engineering and Science. Wiley, New-Jersey 2004.   CrossRef
  15. S. Coles: An Introduction to Statistical Modeling of Extreme Values. Springer, London 2001.   CrossRef
  16. G. Christoph and M. Falk: A note on domains of attraction of p-max stable laws. Stat. Prob. Lett. 28 (1996), 279-284.   DOI:10.1007/BF02110701
  17. A. C. Davison and R. L. Smith: Models for exceedances over high threshold. J. Roy. Stat. Soc. Ser. B. 52 (1990), 393-442.   CrossRef
  18. P. de Zea Bermudez and S. Kotz: Parameter estimation of the generalized Pareto distribution - Part I. J. Stat. Plan. Inf. 140 (2010), 1353-1373.   DOI:10.1016/j.jspi.2008.11.019
  19. B. V. Gnedenko: Sur la distribution limite du terme maximum d'une série aléatoire. Ann. Math. 44 (1943), 423-453.   CrossRef
  20. J. A. Greenwood, J. M. Landwehr, N. C. Matalas and J. R. Wallis: Probability weighted moments: definition and relation to parameters of several distributions expressible in inverse form. Water Resour. Res. 15 (1979), 5, 1049-1054.   DOI:10.1029/WR015i005p01049
  21. J. R. M. Hosking: L-moments: analysis and estimation of distributions using linear combinations of order statistics. J. Roy. Stat. Soc. B52 (1990), 105-124.   CrossRef
  22. J. M. Landwehr, N. C. Matalas and J. R. Wallis: Probability weighted moments compared with some traditional techniques for estimating Gumbel parameters and quantiles. Water Resour. Res. 15 (1979), 1055-1064.   DOI:10.1029/WR015i005p01055
  23. D. Nasri-Roudsari: Limit distributions of generalized order statistics under power normalization. Comm. Statis.- Theory Meth. 28 (1999), 6, 1379-1389.   DOI:10.1080/03610929908832362
  24. E. Pancheva: Limit theorems for extreme order statistics under nonlinear normalization. In: Stability Problems for Stochastic Models, Lecture Notes in Math., Springer, Berlin 1155 (1984), 284-309.   CrossRef
  25. J. Pickands: Statistical inference using extreme order statistics. Ann. Stat. 3 (1975), 119-131.   CrossRef
  26. S. Ravi and T. S. Mavitha: New limit distributions for extreme under a nonlinear normalization. ProbStat Forum 9 (2016), 1-20.   CrossRef