Abstract
This paper contributes to answering a question that is of crucial importance in risk management and extreme value theory: How to select the threshold above which one assumes that the tail of a distribution follows a generalized Pareto distribution. This question has gained increasing attention, particularly in finance institutions, as the recent regulative norms require the assessment of risk at high quantiles. Recent methods answer this question by multiple uses of the standard goodness-of-fit tests. These tests are based on a particular choice of symmetric weighting of the mean square error between the empirical and the fitted tail distributions. Assuming an asymmetric weighting, which rates high quantiles more than small ones, we propose new goodness-of-fit tests and automated threshold selection procedures. We consider a parameterized family of asymmetric weight functions and calculate the corresponding mean square error as a loss function. Then we explicitly determine the risk function as the expected value of the loss function for finite sample. Finally, the risk function can be used to discuss whether a symmetric or asymmetric weight function should be chosen. With this the goodness-of-fit test which should be used in a new method for determining the threshold value is specified.
Similar content being viewed by others
References
Abramowitz M, Stegun IA (eds) (2014) Handbook of mathematical functions with formulas, graphs, and mathematical tables, National Bureau of Standards (Washington DC), Martino Publishing, reprint of 1964 Edition, Mansfield Centre
Aggarwal OP (1955) Some minimax invariant procedures for estimating a cumulative distribution function. Ann Math Stat 26:450–463
Ahmad MI, Sinclair CD, Spurr BD (1988) Assessment of flood frequency models using empirical distribution function statistics. Water Resour Res 24:1323–1328
Anderson TW, Darling DA (1952) Asymptotic theory of certain ’Goodness of Fit’ criteria based on stochastic processes. Ann Math Stat 23:193–212
Anderson TW, Darling DA (1954) A test of goodness of fit. J Am Stat Assoc 49:765–769
Bader B, Yan J, Zhang X (2018) Automated threshold selection for extreme value analysis via Goodness-of-Fit tests with application to batched return level mapping. Ann Appl Stat 12:310–329
Basel Commitee on Banking Supervision (2009) Observed range of practice in key elements of advanced measurement approaches (AMA). Bank for International Settlements, Basel
Chapelle A, Crama Y, Hübner G, Peters JP (2005) Measuring and managing operational risk in financial sector: an integrated framework. Soc Sci Res Netw Electron J. https://doi.org/10.2139/ssrn.675186
Cordeiro GM, Christino CT, Hashimoto EM, Ortega EMM (2013) The beta generalized Rayleigh distribution with application to lifetime data. Stat Pap 54:133–161
Crama Y, Hübner G, Peters JP (2007) Impact of the collection threshold on the determination of the capital charge for operational risk. In: Greg N (ed) Adv Risk Manag. Gregoriou, Palgrave Macmillan, New, York, pp 1–16
Cramér H (1928) On the composition of elementary errors, second paper: statistical applications. Scand Actuar J 1:141–180
Denecke L, Müller ChH (2012) Consistency and robustness of tests and estimators based on depth. J Stat Plan Inference 142:2501–2517
Denecke L, Müller ChH (2014) New robust tests for the parameters of the Weibull distribution for complete and censored data. Metrika 77:585–607
Embrechts P, Klüppelberg C, Mikosch T (2003) Modelling extremal events: for insurance and finance (stochastic modelling and applied probability). Springer, corrected 4th printing, Berlin, Germany
Eugene N, Lee C, Famoye F (2002) Beta-normal distribution and its applications. Commun Stat 31:497–512
Ferguson TS (1967) Math Stat A. Academic Press, New York
Huisman R, Koedijk KG, Kool CJM, Palm F (2001) Tail-Index estimates in small samples. J Bus Econ Stat 19:208–216
Kolmogorov AN (1933) Sulla determinazione empirica di una legge di distribuzione. Giornale dell’Istituto Italiano degli Attuari 4:83–91
Kotz S, Seier E (2009) An analysis of quantile measures of kurtosis: center and tails. Stat Pap 50:553–568
Longin F, Solnik B (2001) Extreme correlation of international equity markets. J Financ 56:649–676
Meintanis SG, Allison J, Santana L (2016) Goodness-of-fit tests for semiparametric and parametric hypotheses based on the probability weighted empirical characteristic function. Stat Pap 57:957–976
Oldham K, Myland J, Spanier J (2009) An atlas of functions, 2nd edn. Springer, New York
Sinclair CD, Spurr BD, Ahmad MI (1990) Modified Anderson Darling test. Commun Stat 19:3677–3686
Smirnov NV (1936) Sur la distribution de \(\omega ^2\)-criterion (critérion de R. von Mises). C R l’Académie Sci Paris 202:449–452
Von Mises RE (1931) Wahrscheinlichkeitsrechnung und ihre Anwendung in der Statistik und theoretischen Physik. Franz Deuticke, Leipzig und Wien
Wagner N, Marsh TA (2004) Tail index estimation in small samples—simulation results for independent and ARCH-type financial return models. Stat Pap 45:545–562
Acknowledgements
The authors would like to thank the editor and the two anonymous referees for their helpful comments.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix A: Auxiliary calculations
Appendix A: Auxiliary calculations
Lemma 1
Let
then
where \(\nu \in {\mathbb {R}}\), \(k\in {\mathbb {N}}\), \((n-1)_{(l)}\) is the Pochhammer notation for falling factorials and \(S_{k,l}\) are the Stirling numbers of the second kind (Abramowitz and Stegun 2014).
Proof
(Lemma 1) To begin, the beta functions \(B(\cdot ,\cdot )\) of Eq. (32) are expressed in terms of the gamma function \(\varGamma (\cdot )\) (Abramowitz and Stegun 2014). Simplifying the fraction yields
Depending on \(\nu \), the possibly resulting poles must be considered and the gamma function should be considered in its analytic continuation \( \varGamma (x+\alpha )= (x)^{(\alpha )}\varGamma (x)\), where \((x)^{(\alpha )}\) is the Pochhammer notation for rising factorials (Abramowitz and Stegun 2014). After simplifying the fraction, we receive
Using the identity \((x)^{(\alpha )} = (x)^{(\beta )} (x+\beta )^{(\alpha - \beta )} \) results in
Remember that \(\frac{\varGamma (n)}{\varGamma (i)} = \left( {\begin{array}{c}n\\ i\end{array}}\right) \frac{i}{n} (n-i)! = \left( {\begin{array}{c}n-1\\ i-1\end{array}}\right) (1)^{(n-i)}\). Then,
Now, \(i^k\) is decomposed into a sum of the falling factorials, where the coefficients consist of Stirling numbers of the second kind \( i^k = \sum _{l=0}^{k} S_{k,l} (i)_{(l)}\). Using the appropriate numbering of the sum with an appropriate extension of the terms gives
By changing the order of the sums and truncating the binomial coefficient, the above equation reduces to
By renumbering the last sum, splitting the term \( (\nu +1)^{(l-1)}\) and using Chu-Vandermonde theorem (Oldham et al. 2009, Ch. 18), this equation reduces to
After renumbering the remaining sum and multiplying the rising factorials, Eq. (32) follows immediately. \(\square \)
Rights and permissions
About this article
Cite this article
Hoffmann, I., Börner, C.J. The risk function of the goodness-of-fit tests for tail models. Stat Papers 62, 1853–1869 (2021). https://doi.org/10.1007/s00362-020-01159-3
Received:
Revised:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00362-020-01159-3