Physics and Chemistry of the Earth

Size: px
Start display at page:

Download "Physics and Chemistry of the Earth"

Transcription

1 Physics and Chemistry of the Earth 34 (2009) Contents lists available at ScienceDirect Physics and Chemistry of the Earth journal homepage: Performances of some parameter estimators of the generalized Pareto distribution over rounded-off samples Roberto Deidda *, Michelangelo Puliga Dipartimento di Ingegneria del Territorio, Facoltá di Ingegneria, Universitá di Cagliari, Piazza d Armi, I Cagliari, Italy article info abstract Article history: Received 25 July 2008 Received in revised form 26 November 2008 Accepted 1 December 2008 Available online 24 December 2008 Keywords: Generalized Pareto distribution Estimators Rounded-off records Root mean square errors Bias Recent analyses on some daily rainfall time series highlighted the presence of records with anomalous rounding (1 and 5 mm), while the standard resolution should be 0.1 or 0.2 mm. Assuming that the generalized Pareto distribution (GPD) can reliably represent the distribution of daily rainfall depths, this study investigates how such discretizations can affect the inference process. The performances of several GPD estimators are compared using the Monte Carlo approach. Synthetic samples are drawn by GPDs with shape and scale parameters in the ranges of values estimated on daily rainfall depth time series. Results show how the relative efficiency of estimators could be very different for continuous or rounded-off samples. Moreover, when the rounding-off magnitude becomes larger than a few millimeters, all the considered estimators reveal very poor performances. Ó 2008 Elsevier Ltd. All rights reserved. 1. Introduction Several statistical distributions have been applied and sometimes specifically proposed to characterize the ordinary and the extreme behaviour of rainfall and discharge observations, as well as other meteo-climatological variables. E.g., the normal, lognormal, exponential, gamma, Pearson Type III, log-pearson Type III, Gumbel, Weibull, two component extreme value (TCEV), generalized extreme value (GEV) and generalized Pareto distribution (GPD) are all distributions which have received consideration in statistical hydrology. Referring the reader to Chow et al. (1988) and Stedinger et al. (1993) for more details on these distributions and their applications in hydrology, we want to stress here the importance of the last two distributions, which appear particularly attractive for their asymptotic statistical properties: namely the GPD and the GEV. Defining the extreme values as the maxima within time blocks (usually assumed a year long in Earth Sciences), it can be proved that, if a limit distribution of these maxima exists, this distribution belongs to the GEV family (Fisher and Tippett, 1928; Gnedenko, 1943). Conversely, when looking at the exceedances above a threshold, it can be proved that the GPD is the expected distribution (Pickands, 1975). Moreover, if a process follows a GPD with a given shape parameter, the block maxima follow a GEV distribution with the same shape parameter (Balkema De Haan Pickands theorem). The reader is referred to Gumbel (1958), Castillo (1988), * Corresponding author. Tel.: ; fax: address: rdeidda@unica.it (R. Deidda). and Coles (2001) and references therein for more details on these important properties of GPD and GEV. The advantage in adopting the couplet of distributions GPD and GEV is thus straightforward. GPD has only two parameters which we can estimate with good accuracy on the large samples of ordinary rainfall time series, while the three GEV parameters must be estimated only on annual maxima and are affected by large estimation errors. Thus one of the reasons why we want to focus on the problem of fitting the GPD on ordinary rainfall time series is that it should be possible, according to theoretical arguments, to estimate first the GPD parameters using a large amount of data and then to use the results to determine the GEV distribution which describes the extremes. Besides these important theoretical features, there are also empirical arguments for adopting the GPD to describe ordinary rainfall time series: an analysis based on the L-moments ratio diagram (Hosking, 1990) performed on 200 daily rainfall time series by Deidda and Puliga (2006) showed that the GPD is the best candidate to describe these datasets. The GPD, first introduced by Pickands (1975) to describe the exceedances over high thresholds, is, because of its nature, sometimes also referred to as points over threshold (or peaks over threshold ) (POT) distribution. Adopting hydrological language, the GPD can describe the partial duration series (PDS) which are obtained by letting a threshold select the highest values from ordinary (continuous) hydrological time series. On the other hand, the block maxima are often referred to as annual maxima series (AMS) because the time block length is usually assumed to be one year. A comparison of the relative performance of the AMS /$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved. doi: /j.pce

2 R. Deidda, M. Puliga / Physics and Chemistry of the Earth 34 (2009) and PDS estimation methods is discussed in Madsen et al. (1997a,b) for rainfall and flood time series: they show that the POT approach in estimating the statistical distribution of rain records for a network of hydrological stations can be an advantageous alternative to block maxima methods. The generalized Pareto distribution GPD has the following equation: ( x u 1=n 1 1 þ n F u ðx; a; nþ ¼PrfX 6 xjx > ug ¼ a n 0 ð1þ 1 exp x u a n ¼ 0 where n is the shape parameter, a the scale parameter, while u is the threshold value. For n ¼ 0 the GPD becomes a simple exponential distribution, for n > 0 is characterized by a heavy tail, while for n < 0 becomes a bounded distribution. The determination of the optimal threshold for GPD fitting is still an open problem. Graphical and numerical methods have been proposed and applied by several authors to detect the threshold (e.g., Davison and Smith, 1990; Smith, 1994; Lang et al., 1999; Dupuis, 1999; Choulakian and Stephens, 2001). More recently, on the basis of a preliminary study on the threshold estimation by Guillou and Hall (2001), a paper by Peng and Qi (2004) proposes an alternative approach to the selection of an optimal threshold: rather than searching for the correct threshold, Peng and Qi (2004) look for powerful and robust estimators with moderated also for too low (or too high) values of the threshold. Deidda and Puliga (2006) highlighted how the presence of rounded-off values in the sample can affect the determination of the optimal threshold, sometimes also masking the presence of the threshold: they proposed to overcome this problem by using a modification of the failure-to-reject method (Choulakian and Stephens, 2001). Deidda and Puliga (2006) suggested computing the regions of acceptance of the goodness of fit tests on roundedoff samples generated via Monte Carlo techniques, adopting the same rounding-off rule as the observed sample. The presence of rounded-off values in rainfall records and the related problems in the inference of probability distributions was then investigated in depth by Deidda (2007): a systematic analysis on 340 daily rainfall time series collected by the rain gauge network of the Sardinian Hydrological Survey (Italy) revealed the presence of significant percentages of roughly rounded off measurements, even at 1 and 5 mm resolutions, rather than at the standard 0.1 or 0.2 mm discretization. Nevertheless, the presence of rounded-off values not only affects the determination of the GPD threshold, but also induces errors in the estimation of the shape n and the scale a parameters. This paper investigates these errors and compares the performances of different estimators when applied to estimate n and a on synthetic samples generated by a GPD with zero threshold and then rounded off at different discretizations. A review of GPD estimators is presented in Section 2. The evaluation of the performances of selected GPD estimators is discussed in Section 3, while the conclusions are drawn in Section A review of GPD estimators Over the last years many estimators have been developed to improve the efficiency and the robustness of the fitting techniques. The estimators can be of several classes: tail index (as Hill or De Haan), maximum likelihood functions, moments, probability weighted moments, medians and goodness of fit based. Every estimator class has some drawbacks and some advantages. The maximum likelihood estimator is based on the maximization of the likelihood function L: this method is very important because it is asymptotically (i.e. for large samples) the best. Nevertheless, when the sample is small or is contaminated by spurious data, the method could provide unrealistic estimates. In a classical study, Hosking and Wallis (1987) compare the performances with those of the methods of moments and of probability weighted moments: they show how the could lead to inaccurate results for small samples. Juárez and Schucany (2004) analyze the performances of the estimator on contaminated GPD samples showing that the method lacks robustness: they introduce a robustified method known as minimum divergence power density estimator. Another estimator class is that of the tail index estimators. The basic idea is to estimate the average slope of the distribution tail using a plot position rule of ranked data in log diagrams. The method is useful for a large variety of distributions like gamma, student and GPD. By combining the tail index with other order statistics it is possible to estimate the GPD parameters as shown in Pickands (1975). The estimator gives generally good performances for large samples of pure data, but must be tuned using a suitable threshold. Moreover, the numerical algorithms for GPD parameters estimation are often affected by convergence problems. The accuracy of the GPD fit depends on the kind of data and thus it must be evaluated for specific cases. In fact, as shown by Rosbjerg et al. (1992), it is possible that, for moderate tails (values of the shape parameter near zero) and small datasets, the description with the ordinary exponential distribution may be more accurate than the GPD one. This is an expected result because the exponential distribution requires the estimation of a single parameter, while the GPD needs the estimation of two parameters. Obviously similar results can be obtained, more generally, by comparing the GPD performances in the case of n known and a unknown with the case of both parameters unknown. Nevertheless, the reader should be careful in adopting these approaches because they are advantageous only if the shape parameter is a priori known. In this study we compare the performances of several estimators of GPD parameters on rounded-off samples. We discarded the estimators based on tail index because we found that the numerical algorithms are very unstable and often fail. The estimators compared here are listed and discussed in the following. For the numerical implementation of the first three widely applied methods the reader is referred to Hosking and Wallis (1987) and Stedinger et al. (1993). For the other (and more recent) methods, the references are provided in the text of each description. Maximum likelihood estimator (). The is a standard and widely adopted estimation technique that can be applied to any statistical distribution. It is based on the idea of finding the set of h parameters which maximize the likelihood function LðX; hþ evaluated on the sample X. We remark that the location parameter u of the GPD can not be obtained by the, in fact the score function ol=ou is unbounded. The maximum likelihood estimates of the remaining shape n and scale a parameters can be obtained in different ways leading to slightly different results: the function can be formalized in the classical bivariate way or in an univariate and computationally more efficient way by introducing a smart substitution for the scale and shape parameters, as suggested by Grimshaw (1993). Many authors have proved that the is the best estimator in the presence of large samples, the asymptotic behaviour of this method is known to be the best possible one. But for small samples (6 100 values) the fit is not always good and the method can be outperformed by other techniques (Hosking and Wallis, 1987), moreover, the numerical algorithms used to estimate the maxima sometimes fail to converge to local maxima, thus robust and powerful computational methods must be used to find the maxima by avoiding convergence problems. Moments estimator (). It represents the simplest method: estimates of the shape n and scale a parameters are

3 628 R. Deidda, M. Puliga / Physics and Chemistry of the Earth 34 (2009) Bias for ξ: all estimators Bias ξ Fig. 1. Bias of shape parameter n for different GPD estimators. The Bias is computed with Monte Carlo techniques over continuos samples of size 500, generated by a GPD with threshold u ¼ 0, a ¼ 7 and n in the range ð 0:5; 0:5Þ. The final result is filtered by a robust gaussian kernel smooth function (Nadaraya, 1964). obtained as simple functions of the mean and the variance of the distribution. The method is theoretically applicable only for values of n < 1=2 because for n! 1=2 the variance tends to be infinite. Hosking and Wallis (1987) suggest using the estimator for n < 1=4. When n 0 the accuracy of the method is close to the estimator. Nevertheless, we must note that the moments method is very sensitive to outliers, in fact the mean and the variance are statistical quantities lacking robustness: a single outlier could dramatically change all these quantities. Probability Weighted Moments estimators ( and ). The probability-weighted moments PWM were introduced by Greenwood et al. (1979) and represent an alternative to the ordinary moments. As for the estimator, parameters can be expressed as a function of PWMs. The PWM estimator is particularly advantageous for small datasets because the probability weighted moments have a smaller uncertainty than the ordinary moments. The best performance is reached for n 0:2, for positive shape values performances are very close to ones, while for n < 0 PWM performances become a little worse than those of. Hosking and Wallis (1987) give two definitions of PWM, uned () and ed (), but the difference can be detected only for small samples. Maximum penalized likelihood (). Coles and Dixon (1999) introduced a weight function for the maximum likelihood function LðX; hþ for n > 0. This estimator corrects the tendency of to diverge for small samples. Minimum density power divergence (). This robust estimator has been introduced by Juárez and Schucany (2004) and was derived from the by using a special function of divergence between the fitted function and the data. A constant is introduced to control the trade-off between robustness and efficiency. This property could be very attractive when dataset are contaminated. Likelihood moment estimator (). This method has recently been proposed by Zhang (2007) as a replacement for the PWM and moments method. This estimator should be efficient and robust but it is slow and computational intensive. for ξ: all estimators ξ Fig. 2. Same as Fig. 1, but for of the shape parameter n.

4 R. Deidda, M. Puliga / Physics and Chemistry of the Earth 34 (2009) Median estimator (). This estimator is the most CPU time intensive, it was designed to be resistant to outliers (Peng and Welsh, 2001). Nevertheless, for pure GPD data its performances are very poor as shown by Juárez and Schucany (2004). Numerical packages containing the implementation of these estimation techniques can be found in the R package POT created and maintained by Ribatet (2007). ξ = 0. α = 7 3. Performances of GPD estimators The performances of a parameter estimator depend on many factors: the sample size, the shape of the sample distribution, the presence of spurious, multivariate or trended (not stable) data, the internal dependence and the algorithmic stability and precision of the estimation technique. For instance the is efficient only for big samples while the PWM estimator is able to estimate ξ = 0. α = 12 ξ = 0.15 α = 7 ξ = 0.15 α = 12 ξ = 0.3 α = 7 ξ = 0.3 α = 12 Fig. 3. Bias for the shape parameter n estimated with different techniques on rounded-off samples. Results are presented as a function of rounding-off magnitude which ranges from 0 to 5 mm. Subplots refer to different couples of shape and scale parameters (see subtitles) selected in the range of representative values of daily time series.

5 630 R. Deidda, M. Puliga / Physics and Chemistry of the Earth 34 (2009) the GPD parameters for small samples, but it has a large for values of n < 0. In order to compare the performances of the estimating techniques described in the previous Section, two groups of tests were carried out using Monte Carlo techniques. The first group aims to compare the performances of estimators on continuous samples, while the second one aims to evaluate the performances on rounded-off records. The performances are evaluated by the Bias and the root mean square error : Bias ¼ Eðh est h true Þ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi h i ¼ E ðh est h true Þ 2 ð2þ ð3þ ξ = 0. α = 7 ξ = 0. α = 12 ξ = 0.15 α = 7 ξ = 0.15 α = 12 ξ = 0.3 α = 7 ξ = 0.3 α = 12 Fig. 4. Same as Fig. 3, but for Bias of the scale parameter a.

6 R. Deidda, M. Puliga / Physics and Chemistry of the Earth 34 (2009) where h est ; h true are the estimated and the true (i.e. used to generate the synthetic samples) values of the parameter respectively. In our case h can be the n and/or the a parameter of the GPD. Bias and are computed on synthetic samples generated by the GPD given by Eq. (1) with threshold u ¼ 0 and ðn; aþ couples of parameters which can be considered representative of daily rainfall records, while the size of each synthetic sample is 500. In addition, in the second group of tests, samples are then rounded off with different discretization magnitudes. Before continuing with the presentation of the results, we highlight that in order to reduce the sampling variability in the computation of the Bias and the, a large number of samples would be necessary for each test (i.e. for each couple of n and a parameters and each discretization). Thus, as a compromise with the computational time required by the amount of carried out tests, on one hand each Monte Carlo simulation was limited to 10,000 synthetic samples, on the other hand a Gaussian kernel smoother (Nadaraya, 1964) was applied to smooth lines plotted in the following Figures, for ξ : rounded data ξ = 0. α = 7 for ξ : rounded data ξ = 0. α = 12 for ξ : rounded data ξ = 0.15 α = 7 for ξ : rounded data ξ = 0.15 α = 12 for ξ : rounded data ξ = 0.3 α = 7 for ξ : rounded data ξ = 0.3 α = 12 Fig. 5. Same as Fig. 3, but for of the shape parameter n.

7 632 R. Deidda, M. Puliga / Physics and Chemistry of the Earth 34 (2009) filtering out the artificial noise due to the residual sampling variability of Bias and estimates Tests over continuous GPD samples In this group of tests, Bias and are computed with Monte Carlo techniques on continuous samples generated by GPD with threshold u ¼ 0, parameters a ¼ 7 and n 2ð 0:5; 0:5Þ. In the plots of the Bias (Fig. 1) and of the (Fig. 2) we can evaluate the performances of the considered methods. We remark again that, at least for large samples, the is expected to be the best estimator, thus lines close to the one give evidence of good performances. For instance, the estimator has a severe breakdown for n > 0:3 and a good accuracy for n 0. The estimator fails for n < 0:3, but this region is of scarce importance in rainfall and flood time series applications since distributions are usually unbounded. The performance of is very good for negative shape values, on the contrary the estimator, as highlighted by Juárez and Schucany (2004), has a poor efficiency (less than 20% of the estimator) and a high Bias: it becomes com- for α : rounded data ξ = 0. α = 7 for α : rounded data ξ = 0. α = 12 for α : rounded data ξ = 0.15 α = 7 for α : rounded data ξ = 0.15 α = 12 for α : rounded data ξ = 0.3 α = 7 for α : rounded data ξ = 0.3 α = 12 Fig. 6. Same as Fig. 3, but for of the scale parameter a.

8 R. Deidda, M. Puliga / Physics and Chemistry of the Earth 34 (2009) petitive only on contaminated samples. The PWM estimators (ed and uned) perform well for n 0:1 where the efficiency is the best possible. Nevertheless below this value the is worse than for the other methods. Finally for the penalized maximum likelihood we have good performances close to ones Tests over rounded-off GPD samples This group of tests investigates the performances of the estimators over rounded-off samples. Recent analyses on daily rainfall time series (Deidda and Puliga, 2006; Deidda, 2007) revealed the presence of rounded-off records with a mixture of different resolutions (0.1, 0.2, 0.5, 1.0 and 5.0 mm). To simplify the interpretation of the results of our analysis, here we consider a single rounding off d for each test. The magnitude of d explores the entire range from 0 to 5 mm with increments of 0.1 mm. We expect that the Bias and the increase with the magnitude of the rounding off. This result could be explained with geometrical arguments: the empirical cumulative distribution function (ecdf) of a rounded-off sample displays a step-like behaviour which produces a big uncertainty in the estimate. To explore the performances of the considered GPD estimators for parameters values in the domain of interest of daily rainfall records, we evaluated the Bias and the for n 2f0; 0:15; 0:30g and a 2f7; 12g that are representative of parameters values estimated over the 200 longest datasets of the Sardinia Region, see e.g. Fig. 5 in Deidda (2007). The Biases in n and a estimates are presented in Figs. 3 and 4, respectively, as a function of the rounding-off resolution: each subplot is obtained by Monte Carlo generations of GPD samples generated with a fixed couple of parameters (n; a), obtained by the combination of the values reported above. The Bias of both n and a estimates increases with the magnitude of the discretization. A similar behaviour is displayed by Figs. 5 and 6, where the of n and a is plotted again versus the rounding-off resolution. The comparison of and Bias for rounded-off samples shows noticeable differences with respect to the continuous case. The estimator has an evident failure in the n estimates for rounding off more than 1 mm while in the continuous case it has very good performances close to (or even better than) the ones. Similar problems affect the PWM estimators: we highlight that the tests on rounded samples are performed with the values n 2f0; 0:15; 0:3g where the PWM performances in the continuous case are very good and close to. Moreover and unexpectedly, looking again at displayed in Figs. 5 and 6, we can observe that the estimator performs better for high n values showing an opposite behaviour with respect to the continuous case where the efficiency becomes worse for large n values. The efficiency in some cases is even better than the one. However, besides the above considerations on the relative performances of a given estimator with respect to the others, we highlight that all the considered methods provide parameter estimates affected by unacceptable errors. In some cases the Bias and the are larger than the 100% of the parameter value of the parent distribution, even when we select the best estimator for the specific case. For instance, looking at the (which is the most significant index because it also includes the Bias), we can observe in Fig. 5 that even for the best estimator the error is often about (or larger than) 0.2 for the n parameter which usually assumes values between 0.0 and 0.3 in hydrological applications. Similar considerations also hold for Fig. 6, where of a is often about 4 mm and more, while a estimates for daily rainfall depths are usually about 10 mm, see again Fig. 5 in Deidda (2007). 4. Conclusions The main objective of this study was the investigation of the performances of some estimators of the GPD shape and scale parameters on rounded-off samples. With this aim, some widely used GPD estimators (such as the maximum likelihood, the moments, and the probability-weighted moments methods) and some other recently proposed ones (such as the maximum penalized likelihood, minimum density power divergence, likelihood moment estimator, median estimator) were considered and compared. Performances were computed with Monte Carlo techniques on synthetic samples generated by GPD with parameter values which can be considered representative of daily rainfall distributions. A first group of tests was performed on continuous samples. Results showed that the maximum likelihood and the maximum penalized likelihood estimators outperform the other estimators. The likelihood moment estimator has a good performance, although slightly worse than the first two previous ones. The simple moment estimator performs quite well for moderated values of the shape parameter (n 0), while the effects of moment divergence become relevant for the shape parameter approaching 0.5. A second group of tests was performed on rounded-off samples, exploring a range of discretizations from 0 to 5 mm detected in daily rainfall time series by Deidda and Puliga (2006) and Deidda (2007). Comparisons among the performances of the considered estimators showed some surprising and unexpected results, revealing that some estimators display a very different behaviour on discretized samples with respect to the continuous case. For instance, the performances of the maximum penalized likelihood estimator slowly worsens on samples rounded at resolutions larger than 1 mm, while the same estimator performs better than the maximum likelihood one for continuous samples. Performances of the probability-weighted moments estimator are very bad with respect to the other estimators, while it shows good performances in the continuous case for positive values of the shape parameter. On the contrary, performances of the moments estimator are relatively better than the other ones, although its relative performances are not good for large positive values of the shape parameter in the continuous case. But in conclusion, although we can speculate on which estimator could perform better than another one for a specific case, it is very important to observe that none of the considered estimators can give acceptable estimates on sample data rounded off at a resolution of a few millimeters. Indeed, computed Bias and are often of the same magnitude as the parameter value to be estimated. Thus the determination of efficient estimators for rounded-off sample is still an open problem. Methods based on multiple threshold fitting and re-parametrization of the estimates, as proposed in Deidda (2007), promise to provide a good solution to this problem. References Castillo, E., Extreme Values Theory in Engineering. Academic Press, San Diego. Choulakian, V., Stephens, M.A., Goodness-of-fit tests for the generalized Pareto distribution. Technometrics 43, Chow, V.T., Maidment, D.R., Mays, L.W., Applied Hydrology. McGraw-Hill, Singapore. Coles, S., An Introduction to Statistical Modeling of Extreme Values. Springer- Verlag, London. Coles, S., Dixon, M., Likelihood-based inference for extreme value models. Extremes 2 (1), Davison, A.C., Smith, R.L., Models for exceedances over high thresholds. J.R. Stat. Soc. B 52 (3), Deidda, R., An efficient rounding-off rule estimator: application to daily rainfall time series. Water Resour. Res. 43, W doi: / 2006WR

9 634 R. Deidda, M. Puliga / Physics and Chemistry of the Earth 34 (2009) Deidda, R., Puliga, M., Sensitivity of goodness of fit statistics to rainfall data rounding off. Phys. Chem. Earth 31 (18), Dupuis, D.J., Exceedances over high thresholds: a guide to threshold selection. Extremes 1 (3), Fisher, R.A., Tippett, L.H., Limiting forms of the frequency distribution of the largest or smallest member of a sample. Proc. Cambridge Philos. Soc. 24, Gnedenko, B.V., Sur la distribution limite du terme maximum d une série aléatoire. Ann. Math. 44, Greenwood, J.A., Landwehr, J.M., Matalas, N.C., Wallis, J.R., Probability weighted moments: definition and relation to parameters of several distributions expressable in inverse form. Water Resour. Res. 15 (5), Grimshaw, S.D., Computing maximum likelihood estimates for the generalized Pareto distribution. Technometrics 35, Guillou, A., Hall, P., A diagnostic for selecting the threshold in extreme value analysis. J. R. Stat. Soc. Ser. B 63, Gumbel, E.J., Statistic of Extremes. Columbia University Press, New York. Hosking, J.R.M., L-moments: Analysis and estimation of distributions using linear combinations of order statistics. J. R. Stat. Soc. Ser. B 52 (2), Hosking, J.R.M., Wallis, J.R., Parameter and quantile estimation for the generalized Pareto distribution. Technometrics 29, Juárez, S., Schucany, W.R., Robust and efficient estimation for the generalized Pareto distribution. Extremes 7 (3), Lang, M., Ouarda, T.B.M.J., Bobée, B., Towards operational guidelines for overthreshold modeling. J. Hydrol. 225, Madsen, H., Rasmussen, P., Rosbjerg, D., 1997a. Comparison of annual maximum series and partial duration series methods for modeling extreme hydrologic events 1. At-site modeling. Water Resour. Res. 33 (4), Madsen, H., Pearson, P., Rosbjerg, D., 1997b. Comparison of annual maximum series and partial duration series methods for modeling extreme hydrologic events 2. Regional modeling. Water Resour. Res. 33 (4), Nadaraya, E.A., On estimating regression. Theory Probabil. Appl. 9 (1), doi: / Peng, L., Qi, Y., Estimating the first and second order parameters of a heavy tailed distribution. Aust. NZ J. Stat. 46 (2), Peng, L., Welsh, A.H., Robust estimation of the generalized Pareto distribution. Extremes 4 (1), Pickands, J., Statistical inference using extreme order statistics. Ann. Stat. 3, Ribatet, M., POT: modelling peaks over a threshold. R. News 7, Rosbjerg, D., Madsen, H., Rasmussen, P.F., Prediction in partial duration series with generalized Pareto-distributed exceedances. Water Resour. Res. 28 (11), Smith, R.L., Multivariate Threshold Methods. Kluwer, Dordrecht. Stedinger, J.R., Vogel, R.M., Foufoula-Georgiou, E., Frequency analysis of extreme events. In: Maidment, D.R. (Ed.), Handbook of Hydrology. McGraw-Hill. Chapter 18. Zhang, J., Likelihood moment estimation for the generalized Pareto distribution. Aust. NZ J. Stat. 49 (1),

How Significant is the BIAS in Low Flow Quantiles Estimated by L- and LH-Moments?

How Significant is the BIAS in Low Flow Quantiles Estimated by L- and LH-Moments? How Significant is the BIAS in Low Flow Quantiles Estimated by L- and LH-Moments? Hewa, G. A. 1, Wang, Q. J. 2, Peel, M. C. 3, McMahon, T. A. 3 and Nathan, R. J. 4 1 University of South Australia, Mawson

More information

Fitting the generalized Pareto distribution to data using maximum goodness-of-fit estimators

Fitting the generalized Pareto distribution to data using maximum goodness-of-fit estimators Computational Statistics & Data Analysis 51 (26) 94 917 www.elsevier.com/locate/csda Fitting the generalized Pareto distribution to data using maximum goodness-of-fit estimators Alberto Luceño E.T.S. de

More information

Extreme Precipitation: An Application Modeling N-Year Return Levels at the Station Level

Extreme Precipitation: An Application Modeling N-Year Return Levels at the Station Level Extreme Precipitation: An Application Modeling N-Year Return Levels at the Station Level Presented by: Elizabeth Shamseldin Joint work with: Richard Smith, Doug Nychka, Steve Sain, Dan Cooley Statistics

More information

Overview of Extreme Value Theory. Dr. Sawsan Hilal space

Overview of Extreme Value Theory. Dr. Sawsan Hilal space Overview of Extreme Value Theory Dr. Sawsan Hilal space Maths Department - University of Bahrain space November 2010 Outline Part-1: Univariate Extremes Motivation Threshold Exceedances Part-2: Bivariate

More information

A class of probability distributions for application to non-negative annual maxima

A class of probability distributions for application to non-negative annual maxima Hydrol. Earth Syst. Sci. Discuss., doi:.94/hess-7-98, 7 A class of probability distributions for application to non-negative annual maxima Earl Bardsley School of Science, University of Waikato, Hamilton

More information

Construction of confidence intervals for extreme rainfall quantiles

Construction of confidence intervals for extreme rainfall quantiles Risk Analysis VIII 93 Construction of confidence intervals for extreme rainfall quantiles A. T. Silva 1, M. M. Portela 1, J. Baez & M. Naghettini 3 1 Instituto Superior Técnico, Portugal Universidad Católica

More information

Regional Estimation from Spatially Dependent Data

Regional Estimation from Spatially Dependent Data Regional Estimation from Spatially Dependent Data R.L. Smith Department of Statistics University of North Carolina Chapel Hill, NC 27599-3260, USA December 4 1990 Summary Regional estimation methods are

More information

Zwiers FW and Kharin VV Changes in the extremes of the climate simulated by CCC GCM2 under CO 2 doubling. J. Climate 11:

Zwiers FW and Kharin VV Changes in the extremes of the climate simulated by CCC GCM2 under CO 2 doubling. J. Climate 11: Statistical Analysis of EXTREMES in GEOPHYSICS Zwiers FW and Kharin VV. 1998. Changes in the extremes of the climate simulated by CCC GCM2 under CO 2 doubling. J. Climate 11:2200 2222. http://www.ral.ucar.edu/staff/ericg/readinggroup.html

More information

Estimation of Generalized Pareto Distribution from Censored Flood Samples using Partial L-moments

Estimation of Generalized Pareto Distribution from Censored Flood Samples using Partial L-moments Estimation of Generalized Pareto Distribution from Censored Flood Samples using Partial L-moments Zahrahtul Amani Zakaria (Corresponding author) Faculty of Informatics, Universiti Sultan Zainal Abidin

More information

A review: regional frequency analysis of annual maximum rainfall in monsoon region of Pakistan using L-moments

A review: regional frequency analysis of annual maximum rainfall in monsoon region of Pakistan using L-moments International Journal of Advanced Statistics and Probability, 1 (3) (2013) 97-101 Science Publishing Corporation www.sciencepubco.com/index.php/ijasp A review: regional frequency analysis of annual maximum

More information

PUBLICATIONS. Water Resources Research

PUBLICATIONS. Water Resources Research PUBLICATIONS Water Resources Research RESEARCH ARTICLE Key Points: Critical review of representative methods for GP (generalized Pareto) threshold detection Application to 1714 overcentennial daily rainfall

More information

Journal of Biostatistics and Epidemiology

Journal of Biostatistics and Epidemiology Journal of Biostatistics and Epidemiology Original Article Robust correlation coefficient goodness-of-fit test for the Gumbel distribution Abbas Mahdavi 1* 1 Department of Statistics, School of Mathematical

More information

EXTREMAL QUANTILES OF MAXIMUMS FOR STATIONARY SEQUENCES WITH PSEUDO-STATIONARY TREND WITH APPLICATIONS IN ELECTRICITY CONSUMPTION ALEXANDR V.

EXTREMAL QUANTILES OF MAXIMUMS FOR STATIONARY SEQUENCES WITH PSEUDO-STATIONARY TREND WITH APPLICATIONS IN ELECTRICITY CONSUMPTION ALEXANDR V. MONTENEGRIN STATIONARY JOURNAL TREND WITH OF ECONOMICS, APPLICATIONS Vol. IN 9, ELECTRICITY No. 4 (December CONSUMPTION 2013), 53-63 53 EXTREMAL QUANTILES OF MAXIMUMS FOR STATIONARY SEQUENCES WITH PSEUDO-STATIONARY

More information

Extreme Value Analysis and Spatial Extremes

Extreme Value Analysis and Spatial Extremes Extreme Value Analysis and Department of Statistics Purdue University 11/07/2013 Outline Motivation 1 Motivation 2 Extreme Value Theorem and 3 Bayesian Hierarchical Models Copula Models Max-stable Models

More information

Introduction to Algorithmic Trading Strategies Lecture 10

Introduction to Algorithmic Trading Strategies Lecture 10 Introduction to Algorithmic Trading Strategies Lecture 10 Risk Management Haksun Li haksun.li@numericalmethod.com www.numericalmethod.com Outline Value at Risk (VaR) Extreme Value Theory (EVT) References

More information

Abstract: In this short note, I comment on the research of Pisarenko et al. (2014) regarding the

Abstract: In this short note, I comment on the research of Pisarenko et al. (2014) regarding the Comment on Pisarenko et al. Characterization of the Tail of the Distribution of Earthquake Magnitudes by Combining the GEV and GPD Descriptions of Extreme Value Theory Mathias Raschke Institution: freelancer

More information

Usefulness of the reversible jump Markov chain Monte Carlo model in regional flood frequency analysis

Usefulness of the reversible jump Markov chain Monte Carlo model in regional flood frequency analysis Click Here for Full Article WATER RESOURCES RESEARCH, VOL. 43,, doi:10.1029/2006wr005525, 2007 Usefulness of the reversible jump Markov chain Monte Carlo model in regional flood frequency analysis M. Ribatet,

More information

Richard L. Smith Department of Statistics and Operations Research University of North Carolina Chapel Hill, NC

Richard L. Smith Department of Statistics and Operations Research University of North Carolina Chapel Hill, NC EXTREME VALUE THEORY Richard L. Smith Department of Statistics and Operations Research University of North Carolina Chapel Hill, NC 27599-3260 rls@email.unc.edu AMS Committee on Probability and Statistics

More information

LQ-Moments for Statistical Analysis of Extreme Events

LQ-Moments for Statistical Analysis of Extreme Events Journal of Modern Applied Statistical Methods Volume 6 Issue Article 5--007 LQ-Moments for Statistical Analysis of Extreme Events Ani Shabri Universiti Teknologi Malaysia Abdul Aziz Jemain Universiti Kebangsaan

More information

Robust and Efficient Estimation for the Generalized Pareto Distribution

Robust and Efficient Estimation for the Generalized Pareto Distribution Robust and Efficient Estimation for the Generalized Pareto Distribution Sergio F. Juárez Faculty of Statistics and Informatics Veracruzana University, Xalapa, Ver, México email: sejuarez@uv.mx and William

More information

On the Application of the Generalized Pareto Distribution for Statistical Extrapolation in the Assessment of Dynamic Stability in Irregular Waves

On the Application of the Generalized Pareto Distribution for Statistical Extrapolation in the Assessment of Dynamic Stability in Irregular Waves On the Application of the Generalized Pareto Distribution for Statistical Extrapolation in the Assessment of Dynamic Stability in Irregular Waves Bradley Campbell 1, Vadim Belenky 1, Vladas Pipiras 2 1.

More information

ON THE TWO STEP THRESHOLD SELECTION FOR OVER-THRESHOLD MODELLING

ON THE TWO STEP THRESHOLD SELECTION FOR OVER-THRESHOLD MODELLING ON THE TWO STEP THRESHOLD SELECTION FOR OVER-THRESHOLD MODELLING Pietro Bernardara (1,2), Franck Mazas (3), Jérôme Weiss (1,2), Marc Andreewsky (1), Xavier Kergadallan (4), Michel Benoît (1,2), Luc Hamm

More information

L-momenty s rušivou regresí

L-momenty s rušivou regresí L-momenty s rušivou regresí Jan Picek, Martin Schindler e-mail: jan.picek@tul.cz TECHNICKÁ UNIVERZITA V LIBERCI ROBUST 2016 J. Picek, M. Schindler, TUL L-momenty s rušivou regresí 1/26 Motivation 1 Development

More information

Generalized fiducial confidence intervals for extremes

Generalized fiducial confidence intervals for extremes Extremes (2012) 15:67 87 DOI 10.1007/s10687-011-0127-9 Generalized fiducial confidence intervals for extremes Damian V. Wandler Jan Hannig Received: 1 December 2009 / Revised: 13 December 2010 / Accepted:

More information

Regional Frequency Analysis of Extreme Climate Events. Theoretical part of REFRAN-CV

Regional Frequency Analysis of Extreme Climate Events. Theoretical part of REFRAN-CV Regional Frequency Analysis of Extreme Climate Events. Theoretical part of REFRAN-CV Course outline Introduction L-moment statistics Identification of Homogeneous Regions L-moment ratio diagrams Example

More information

Does k-th Moment Exist?

Does k-th Moment Exist? Does k-th Moment Exist? Hitomi, K. 1 and Y. Nishiyama 2 1 Kyoto Institute of Technology, Japan 2 Institute of Economic Research, Kyoto University, Japan Email: hitomi@kit.ac.jp Keywords: Existence of moments,

More information

Battle of extreme value distributions: A global survey on extreme daily rainfall

Battle of extreme value distributions: A global survey on extreme daily rainfall WATER RESOURCES RESEARCH, VOL. 49, 187 201, doi:10.1029/2012wr012557, 2013 Battle of extreme value distributions: A global survey on extreme daily rainfall Simon Michael Papalexiou, 1 and Demetris Koutsoyiannis

More information

The Goodness-of-fit Test for Gumbel Distribution: A Comparative Study

The Goodness-of-fit Test for Gumbel Distribution: A Comparative Study MATEMATIKA, 2012, Volume 28, Number 1, 35 48 c Department of Mathematics, UTM. The Goodness-of-fit Test for Gumbel Distribution: A Comparative Study 1 Nahdiya Zainal Abidin, 2 Mohd Bakri Adam and 3 Habshah

More information

Overview of Extreme Value Analysis (EVA)

Overview of Extreme Value Analysis (EVA) Overview of Extreme Value Analysis (EVA) Brian Reich North Carolina State University July 26, 2016 Rossbypalooza Chicago, IL Brian Reich Overview of Extreme Value Analysis (EVA) 1 / 24 Importance of extremes

More information

Regional estimation of rainfall intensity-duration-frequency curves using generalized least squares regression of partial duration series statistics

Regional estimation of rainfall intensity-duration-frequency curves using generalized least squares regression of partial duration series statistics WATER RESOURCES RESEARCH, VOL. 38, NO. 11, 1239, doi:10.1029/2001wr001125, 2002 Regional estimation of rainfall intensity-duration-frequency curves using generalized least squares regression of partial

More information

Financial Econometrics and Volatility Models Extreme Value Theory

Financial Econometrics and Volatility Models Extreme Value Theory Financial Econometrics and Volatility Models Extreme Value Theory Eric Zivot May 3, 2010 1 Lecture Outline Modeling Maxima and Worst Cases The Generalized Extreme Value Distribution Modeling Extremes Over

More information

Wei-han Liu Department of Banking and Finance Tamkang University. R/Finance 2009 Conference 1

Wei-han Liu Department of Banking and Finance Tamkang University. R/Finance 2009 Conference 1 Detecting Structural Breaks in Tail Behavior -From the Perspective of Fitting the Generalized Pareto Distribution Wei-han Liu Department of Banking and Finance Tamkang University R/Finance 2009 Conference

More information

Bayesian Modelling of Extreme Rainfall Data

Bayesian Modelling of Extreme Rainfall Data Bayesian Modelling of Extreme Rainfall Data Elizabeth Smith A thesis submitted for the degree of Doctor of Philosophy at the University of Newcastle upon Tyne September 2005 UNIVERSITY OF NEWCASTLE Bayesian

More information

A TEST OF FIT FOR THE GENERALIZED PARETO DISTRIBUTION BASED ON TRANSFORMS

A TEST OF FIT FOR THE GENERALIZED PARETO DISTRIBUTION BASED ON TRANSFORMS A TEST OF FIT FOR THE GENERALIZED PARETO DISTRIBUTION BASED ON TRANSFORMS Dimitrios Konstantinides, Simos G. Meintanis Department of Statistics and Acturial Science, University of the Aegean, Karlovassi,

More information

Tests of the Generalized Pareto Distribution for Predicting Extreme Wind Speeds

Tests of the Generalized Pareto Distribution for Predicting Extreme Wind Speeds SEPTEMBER 2000 BRABSON AND PALUTIKOF 1627 Tests of the Generalized Pareto Distribution for Predicting Extreme Wind Speeds B. B. BRABSON Department of Physics, Indiana University, Bloomington, Indiana J.

More information

A MODIFICATION OF HILL S TAIL INDEX ESTIMATOR

A MODIFICATION OF HILL S TAIL INDEX ESTIMATOR L. GLAVAŠ 1 J. JOCKOVIĆ 2 A MODIFICATION OF HILL S TAIL INDEX ESTIMATOR P. MLADENOVIĆ 3 1, 2, 3 University of Belgrade, Faculty of Mathematics, Belgrade, Serbia Abstract: In this paper, we study a class

More information

The battle of extreme value distributions: A global survey on the extreme

The battle of extreme value distributions: A global survey on the extreme 1 2 The battle of extreme value distributions: A global survey on the extreme daily rainfall 3 Simon Michael Papalexiou and Demetris Koutsoyiannis 4 5 Department of Water Resources, Faculty of Civil Engineering,

More information

Discussion on Human life is unlimited but short by Holger Rootzén and Dmitrii Zholud

Discussion on Human life is unlimited but short by Holger Rootzén and Dmitrii Zholud Extremes (2018) 21:405 410 https://doi.org/10.1007/s10687-018-0322-z Discussion on Human life is unlimited but short by Holger Rootzén and Dmitrii Zholud Chen Zhou 1 Received: 17 April 2018 / Accepted:

More information

MFM Practitioner Module: Quantitiative Risk Management. John Dodson. October 14, 2015

MFM Practitioner Module: Quantitiative Risk Management. John Dodson. October 14, 2015 MFM Practitioner Module: Quantitiative Risk Management October 14, 2015 The n-block maxima 1 is a random variable defined as M n max (X 1,..., X n ) for i.i.d. random variables X i with distribution function

More information

Bayesian GLS for Regionalization of Flood Characteristics in Korea

Bayesian GLS for Regionalization of Flood Characteristics in Korea Bayesian GLS for Regionalization of Flood Characteristics in Korea Dae Il Jeong 1, Jery R. Stedinger 2, Young-Oh Kim 3, and Jang Hyun Sung 4 1 Post-doctoral Fellow, School of Civil and Environmental Engineering,

More information

Estimation of risk measures for extreme pluviometrical measurements

Estimation of risk measures for extreme pluviometrical measurements Estimation of risk measures for extreme pluviometrical measurements by Jonathan EL METHNI in collaboration with Laurent GARDES & Stéphane GIRARD 26th Annual Conference of The International Environmetrics

More information

IT S TIME FOR AN UPDATE EXTREME WAVES AND DIRECTIONAL DISTRIBUTIONS ALONG THE NEW SOUTH WALES COASTLINE

IT S TIME FOR AN UPDATE EXTREME WAVES AND DIRECTIONAL DISTRIBUTIONS ALONG THE NEW SOUTH WALES COASTLINE IT S TIME FOR AN UPDATE EXTREME WAVES AND DIRECTIONAL DISTRIBUTIONS ALONG THE NEW SOUTH WALES COASTLINE M Glatz 1, M Fitzhenry 2, M Kulmar 1 1 Manly Hydraulics Laboratory, Department of Finance, Services

More information

Modelação de valores extremos e sua importância na

Modelação de valores extremos e sua importância na Modelação de valores extremos e sua importância na segurança e saúde Margarida Brito Departamento de Matemática FCUP (FCUP) Valores Extremos - DemSSO 1 / 12 Motivation Consider the following events Occurance

More information

Efficient Estimation of Distributional Tail Shape and the Extremal Index with Applications to Risk Management

Efficient Estimation of Distributional Tail Shape and the Extremal Index with Applications to Risk Management Journal of Mathematical Finance, 2016, 6, 626-659 http://www.scirp.org/journal/jmf ISSN Online: 2162-2442 ISSN Print: 2162-2434 Efficient Estimation of Distributional Tail Shape and the Extremal Index

More information

Lecture 2 APPLICATION OF EXREME VALUE THEORY TO CLIMATE CHANGE. Rick Katz

Lecture 2 APPLICATION OF EXREME VALUE THEORY TO CLIMATE CHANGE. Rick Katz 1 Lecture 2 APPLICATION OF EXREME VALUE THEORY TO CLIMATE CHANGE Rick Katz Institute for Study of Society and Environment National Center for Atmospheric Research Boulder, CO USA email: rwk@ucar.edu Home

More information

The use of L-moments for regionalizing flow records in the Rio Uruguai basin: a case study

The use of L-moments for regionalizing flow records in the Rio Uruguai basin: a case study Regionalization in Ifylwltm (Proceedings of the Ljubljana Symposium, April 1990). IAHS Publ. no. 191, 1990. The use of L-moments for regionalizing flow records in the Rio Uruguai basin: a case study ROBM

More information

Effect of trends on the estimation of extreme precipitation quantiles

Effect of trends on the estimation of extreme precipitation quantiles Hydrology Days 2010 Effect of trends on the estimation of extreme precipitation quantiles Antonino Cancelliere, Brunella Bonaccorso, Giuseppe Rossi Department of Civil and Environmental Engineering, University

More information

Sharp statistical tools Statistics for extremes

Sharp statistical tools Statistics for extremes Sharp statistical tools Statistics for extremes Georg Lindgren Lund University October 18, 2012 SARMA Background Motivation We want to predict outside the range of observations Sums, averages and proportions

More information

Bayesian nonparametrics for multivariate extremes including censored data. EVT 2013, Vimeiro. Anne Sabourin. September 10, 2013

Bayesian nonparametrics for multivariate extremes including censored data. EVT 2013, Vimeiro. Anne Sabourin. September 10, 2013 Bayesian nonparametrics for multivariate extremes including censored data Anne Sabourin PhD advisors: Anne-Laure Fougères (Lyon 1), Philippe Naveau (LSCE, Saclay). Joint work with Benjamin Renard, IRSTEA,

More information

LITERATURE REVIEW. History. In 1888, the U.S. Signal Service installed the first automatic rain gage used to

LITERATURE REVIEW. History. In 1888, the U.S. Signal Service installed the first automatic rain gage used to LITERATURE REVIEW History In 1888, the U.S. Signal Service installed the first automatic rain gage used to record intensive precipitation for short periods (Yarnell, 1935). Using the records from this

More information

Frequency Estimation of Rare Events by Adaptive Thresholding

Frequency Estimation of Rare Events by Adaptive Thresholding Frequency Estimation of Rare Events by Adaptive Thresholding J. R. M. Hosking IBM Research Division 2009 IBM Corporation Motivation IBM Research When managing IT systems, there is a need to identify transactions

More information

PUBLICATIONS. Water Resources Research. A modified weighted function method for parameter estimation of Pearson type three distribution

PUBLICATIONS. Water Resources Research. A modified weighted function method for parameter estimation of Pearson type three distribution PUBLICATIONS Water Resources Research RESEARCH ARTICLE Key Points: Modified weighted function method was introduced Monte-Carlo experiment was carried out to simulate a large number of samples New method

More information

Spatial and temporal extremes of wildfire sizes in Portugal ( )

Spatial and temporal extremes of wildfire sizes in Portugal ( ) International Journal of Wildland Fire 2009, 18, 983 991. doi:10.1071/wf07044_ac Accessory publication Spatial and temporal extremes of wildfire sizes in Portugal (1984 2004) P. de Zea Bermudez A, J. Mendes

More information

Challenges in implementing worst-case analysis

Challenges in implementing worst-case analysis Challenges in implementing worst-case analysis Jon Danielsson Systemic Risk Centre, lse,houghton Street, London WC2A 2AE, UK Lerby M. Ergun Systemic Risk Centre, lse,houghton Street, London WC2A 2AE, UK

More information

Maximum Monthly Rainfall Analysis Using L-Moments for an Arid Region in Isfahan Province, Iran

Maximum Monthly Rainfall Analysis Using L-Moments for an Arid Region in Isfahan Province, Iran 494 J O U R N A L O F A P P L I E D M E T E O R O L O G Y A N D C L I M A T O L O G Y VOLUME 46 Maximum Monthly Rainfall Analysis Using L-Moments for an Arid Region in Isfahan Province, Iran S. SAEID ESLAMIAN*

More information

Influence of parameter estimation uncertainty in Kriging: Part 2 Test and case study applications

Influence of parameter estimation uncertainty in Kriging: Part 2 Test and case study applications Hydrology and Earth System Influence Sciences, of 5(), parameter 5 3 estimation (1) uncertainty EGS in Kriging: Part Test and case study applications Influence of parameter estimation uncertainty in Kriging:

More information

Estimation of spatial max-stable models using threshold exceedances

Estimation of spatial max-stable models using threshold exceedances Estimation of spatial max-stable models using threshold exceedances arxiv:1205.1107v1 [stat.ap] 5 May 2012 Jean-Noel Bacro I3M, Université Montpellier II and Carlo Gaetan DAIS, Università Ca Foscari -

More information

Journal of Environmental Statistics

Journal of Environmental Statistics jes Journal of Environmental Statistics February 2010, Volume 1, Issue 3. http://www.jenvstat.org Exponentiated Gumbel Distribution for Estimation of Return Levels of Significant Wave Height Klara Persson

More information

APPLICATION OF EXTREMAL THEORY TO THE PRECIPITATION SERIES IN NORTHERN MORAVIA

APPLICATION OF EXTREMAL THEORY TO THE PRECIPITATION SERIES IN NORTHERN MORAVIA APPLICATION OF EXTREMAL THEORY TO THE PRECIPITATION SERIES IN NORTHERN MORAVIA DANIELA JARUŠKOVÁ Department of Mathematics, Czech Technical University, Prague; jarus@mat.fsv.cvut.cz 1. Introduction The

More information

Peaks-Over-Threshold Modelling of Environmental Data

Peaks-Over-Threshold Modelling of Environmental Data U.U.D.M. Project Report 2014:33 Peaks-Over-Threshold Modelling of Environmental Data Esther Bommier Examensarbete i matematik, 30 hp Handledare och examinator: Jesper Rydén September 2014 Department of

More information

Small sample corrections for LTS and MCD

Small sample corrections for LTS and MCD Metrika (2002) 55: 111 123 > Springer-Verlag 2002 Small sample corrections for LTS and MCD G. Pison, S. Van Aelst*, and G. Willems Department of Mathematics and Computer Science, Universitaire Instelling

More information

Package homtest. February 20, 2015

Package homtest. February 20, 2015 Version 1.0-5 Date 2009-03-26 Package homtest February 20, 2015 Title Homogeneity tests for Regional Frequency Analysis Author Alberto Viglione Maintainer Alberto Viglione

More information

Automated, Efficient, and Practical Extreme Value Analysis with Environmental Applications

Automated, Efficient, and Practical Extreme Value Analysis with Environmental Applications Automated, Efficient, and Practical Extreme Value Analysis with Environmental Applications arxiv:1611.08261v1 [stat.me] 24 Nov 2016 Brian M. Bader, Ph.D. University of Connecticut, 2016 ABSTRACT Although

More information

Contributions to extreme-value analysis

Contributions to extreme-value analysis Contributions to extreme-value analysis Stéphane Girard INRIA Rhône-Alpes & LJK (team MISTIS). 655, avenue de l Europe, Montbonnot. 38334 Saint-Ismier Cedex, France Stephane.Girard@inria.fr February 6,

More information

International Journal of World Research, Vol - 1, Issue - XVI, April 2015 Print ISSN: X

International Journal of World Research, Vol - 1, Issue - XVI, April 2015 Print ISSN: X (1) ESTIMATION OF MAXIMUM FLOOD DISCHARGE USING GAMMA AND EXTREME VALUE FAMILY OF PROBABILITY DISTRIBUTIONS N. Vivekanandan Assistant Research Officer Central Water and Power Research Station, Pune, India

More information

HIERARCHICAL MODELS IN EXTREME VALUE THEORY

HIERARCHICAL MODELS IN EXTREME VALUE THEORY HIERARCHICAL MODELS IN EXTREME VALUE THEORY Richard L. Smith Department of Statistics and Operations Research, University of North Carolina, Chapel Hill and Statistical and Applied Mathematical Sciences

More information

MAXIMUM WIND GUST RETURN PERIODS FOR OKLAHOMA USING THE OKLAHOMA MESONET. Andrew J. Reader Oklahoma Climatological Survey, Norman, OK. 2.

MAXIMUM WIND GUST RETURN PERIODS FOR OKLAHOMA USING THE OKLAHOMA MESONET. Andrew J. Reader Oklahoma Climatological Survey, Norman, OK. 2. J3.14 MAXIMUM WIND GUST RETURN PERIODS FOR OKLAHOMA USING THE OKLAHOMA MESONET Andrew J. Reader Oklahoma Climatological Survey, Norman, OK 1. Introduction It is well known that Oklahoma experiences convective

More information

Extreme Value Theory and Applications

Extreme Value Theory and Applications Extreme Value Theory and Deauville - 04/10/2013 Extreme Value Theory and Introduction Asymptotic behavior of the Sum Extreme (from Latin exter, exterus, being on the outside) : Exceeding the ordinary,

More information

Bivariate generalized Pareto distribution

Bivariate generalized Pareto distribution Bivariate generalized Pareto distribution in practice Eötvös Loránd University, Budapest, Hungary Minisymposium on Uncertainty Modelling 27 September 2011, CSASC 2011, Krems, Austria Outline Short summary

More information

Generalized maximum-likelihood generalized extreme-value quantile estimators for hydrologic data

Generalized maximum-likelihood generalized extreme-value quantile estimators for hydrologic data WATER RESOURCES RESEARCH, VOL. 36, NO. 3, PAGES 737-744, MARCH 2000 Generalized maximum-likelihood generalized extreme-value quantile estimators for hydrologic data Eduardo S. Martins FUNCEME-Governo do

More information

R.Garçon, F.Garavaglia, J.Gailhard, E.Paquet, F.Gottardi EDF-DTG

R.Garçon, F.Garavaglia, J.Gailhard, E.Paquet, F.Gottardi EDF-DTG Homogeneous samples and reliability of probabilistic models : using an atmospheric circulation patterns sampling for a better estimation of extreme rainfall probability R.Garçon, F.Garavaglia, J.Gailhard,

More information

EXTREMAL MODELS AND ENVIRONMENTAL APPLICATIONS. Rick Katz

EXTREMAL MODELS AND ENVIRONMENTAL APPLICATIONS. Rick Katz 1 EXTREMAL MODELS AND ENVIRONMENTAL APPLICATIONS Rick Katz Institute for Study of Society and Environment National Center for Atmospheric Research Boulder, CO USA email: rwk@ucar.edu Home page: www.isse.ucar.edu/hp_rick/

More information

RISK AND EXTREMES: ASSESSING THE PROBABILITIES OF VERY RARE EVENTS

RISK AND EXTREMES: ASSESSING THE PROBABILITIES OF VERY RARE EVENTS RISK AND EXTREMES: ASSESSING THE PROBABILITIES OF VERY RARE EVENTS Richard L. Smith Department of Statistics and Operations Research University of North Carolina Chapel Hill, NC 27599-3260 rls@email.unc.edu

More information

The utility of L-moment ratio diagrams for selecting a regional probability distribution

The utility of L-moment ratio diagrams for selecting a regional probability distribution Hydrological Sciences Journal ISSN: 0262-6667 (Print) 250-3435 (Online) Journal homepage: http://www.tandfonline.com/loi/thsj20 The utility of L-moment ratio diagrams for selecting a regional probability

More information

Modified Kolmogorov-Smirnov Test of Goodness of Fit. Catalonia-BarcelonaTECH, Spain

Modified Kolmogorov-Smirnov Test of Goodness of Fit. Catalonia-BarcelonaTECH, Spain 152/304 CoDaWork 2017 Abbadia San Salvatore (IT) Modified Kolmogorov-Smirnov Test of Goodness of Fit G.S. Monti 1, G. Mateu-Figueras 2, M. I. Ortego 3, V. Pawlowsky-Glahn 2 and J. J. Egozcue 3 1 Department

More information

INCORPORATION OF WEIBULL DISTRIBUTION IN L-MOMENTS METHOD FOR REGIONAL FREQUENCY ANALYSIS OF PEAKS-OVER-THRESHOLD WAVE HEIGHTS

INCORPORATION OF WEIBULL DISTRIBUTION IN L-MOMENTS METHOD FOR REGIONAL FREQUENCY ANALYSIS OF PEAKS-OVER-THRESHOLD WAVE HEIGHTS INCORPORATION OF WEIBULL DISTRIBUTION IN L-MOMENTS METHOD FOR REGIONAL FREQUENCY ANALYSIS OF PEAKS-OVER-THRESHOLD WAVE HEIGHTS Yoshimi Goda, Masanobu Kudaa, and Hiroyasu Kawai The L-moments of the distribution

More information

On the modelling of extreme droughts

On the modelling of extreme droughts Modelling and Management of Sustainable Basin-scale Water Resource Systems (Proceedings of a Boulder Symposium, July 1995). IAHS Publ. no. 231, 1995. 377 _ On the modelling of extreme droughts HENRIK MADSEN

More information

A NOTE ON SECOND ORDER CONDITIONS IN EXTREME VALUE THEORY: LINKING GENERAL AND HEAVY TAIL CONDITIONS

A NOTE ON SECOND ORDER CONDITIONS IN EXTREME VALUE THEORY: LINKING GENERAL AND HEAVY TAIL CONDITIONS REVSTAT Statistical Journal Volume 5, Number 3, November 2007, 285 304 A NOTE ON SECOND ORDER CONDITIONS IN EXTREME VALUE THEORY: LINKING GENERAL AND HEAVY TAIL CONDITIONS Authors: M. Isabel Fraga Alves

More information

COMPARISON OF THE ESTIMATORS OF THE LOCATION AND SCALE PARAMETERS UNDER THE MIXTURE AND OUTLIER MODELS VIA SIMULATION

COMPARISON OF THE ESTIMATORS OF THE LOCATION AND SCALE PARAMETERS UNDER THE MIXTURE AND OUTLIER MODELS VIA SIMULATION (REFEREED RESEARCH) COMPARISON OF THE ESTIMATORS OF THE LOCATION AND SCALE PARAMETERS UNDER THE MIXTURE AND OUTLIER MODELS VIA SIMULATION Hakan S. Sazak 1, *, Hülya Yılmaz 2 1 Ege University, Department

More information

for explaining hydrological losses in South Australian catchments by S. H. P. W. Gamage The Cryosphere

for explaining hydrological losses in South Australian catchments by S. H. P. W. Gamage The Cryosphere Geoscientific Model Development pen Access Geoscientific Model Development pen Access Hydrology and Hydrol. Earth Syst. Sci. Discuss., 10, C2196 C2210, 2013 www.hydrol-earth-syst-sci-discuss.net/10/c2196/2013/

More information

PLANNED UPGRADE OF NIWA S HIGH INTENSITY RAINFALL DESIGN SYSTEM (HIRDS)

PLANNED UPGRADE OF NIWA S HIGH INTENSITY RAINFALL DESIGN SYSTEM (HIRDS) PLANNED UPGRADE OF NIWA S HIGH INTENSITY RAINFALL DESIGN SYSTEM (HIRDS) G.A. Horrell, C.P. Pearson National Institute of Water and Atmospheric Research (NIWA), Christchurch, New Zealand ABSTRACT Statistics

More information

Estimating return levels from maxima of non-stationary random sequences using the Generalized PWM method

Estimating return levels from maxima of non-stationary random sequences using the Generalized PWM method Nonlin. Processes Geophys., 15, 1033 1039, 2008 Authors 2008. This work is distributed under the Creative Commons Attribution 3.0 License. Nonlinear Processes in Geophysics Estimating return levels from

More information

Bayesian Point Process Modeling for Extreme Value Analysis, with an Application to Systemic Risk Assessment in Correlated Financial Markets

Bayesian Point Process Modeling for Extreme Value Analysis, with an Application to Systemic Risk Assessment in Correlated Financial Markets Bayesian Point Process Modeling for Extreme Value Analysis, with an Application to Systemic Risk Assessment in Correlated Financial Markets Athanasios Kottas Department of Applied Mathematics and Statistics,

More information

ISSN: (Print) (Online) Journal homepage:

ISSN: (Print) (Online) Journal homepage: Hydrological Sciences Journal ISSN: 0262-6667 (Print) 2150-3435 (Online) Journal homepage: http://www.tandfonline.com/loi/thsj20 he use of resampling for estimating confidence intervals for single site

More information

WEIGHTED QUANTILE REGRESSION THEORY AND ITS APPLICATION. Abstract

WEIGHTED QUANTILE REGRESSION THEORY AND ITS APPLICATION. Abstract Journal of Data Science,17(1). P. 145-160,2019 DOI:10.6339/JDS.201901_17(1).0007 WEIGHTED QUANTILE REGRESSION THEORY AND ITS APPLICATION Wei Xiong *, Maozai Tian 2 1 School of Statistics, University of

More information

APPROXIMATING THE GENERALIZED BURR-GAMMA WITH A GENERALIZED PARETO-TYPE OF DISTRIBUTION A. VERSTER AND D.J. DE WAAL ABSTRACT

APPROXIMATING THE GENERALIZED BURR-GAMMA WITH A GENERALIZED PARETO-TYPE OF DISTRIBUTION A. VERSTER AND D.J. DE WAAL ABSTRACT APPROXIMATING THE GENERALIZED BURR-GAMMA WITH A GENERALIZED PARETO-TYPE OF DISTRIBUTION A. VERSTER AND D.J. DE WAAL ABSTRACT In this paper the Generalized Burr-Gamma (GBG) distribution is considered to

More information

Review of existing statistical methods for flood frequency estimation in Greece

Review of existing statistical methods for flood frequency estimation in Greece EU COST Action ES0901: European Procedures for Flood Frequency Estimation (FloodFreq) 3 rd Management Committee Meeting, Prague, 28 29 October 2010 WG2: Assessment of statistical methods for flood frequency

More information

Monte Carlo Studies. The response in a Monte Carlo study is a random variable.

Monte Carlo Studies. The response in a Monte Carlo study is a random variable. Monte Carlo Studies The response in a Monte Carlo study is a random variable. The response in a Monte Carlo study has a variance that comes from the variance of the stochastic elements in the data-generating

More information

Regression Analysis for Data Containing Outliers and High Leverage Points

Regression Analysis for Data Containing Outliers and High Leverage Points Alabama Journal of Mathematics 39 (2015) ISSN 2373-0404 Regression Analysis for Data Containing Outliers and High Leverage Points Asim Kumer Dey Department of Mathematics Lamar University Md. Amir Hossain

More information

A world-wide investigation of the probability distribution of daily rainfall

A world-wide investigation of the probability distribution of daily rainfall International Precipitation Conference (IPC10) Coimbra, Portugal, 23 25 June 2010 Topic 1 Extreme precipitation events: Physics- and statistics-based descriptions A world-wide investigation of the probability

More information

CHARACTERIZATION OF THE TAIL OF RIVER FLOW DATA BY GENERALIZED PARETO DISTRIBUTION

CHARACTERIZATION OF THE TAIL OF RIVER FLOW DATA BY GENERALIZED PARETO DISTRIBUTION Journal of Statistical Research 2016, Vol. 48-50, No. 2, pp. 55-70 ISSN 0256-422 X CHARACTERIZATION OF THE TAIL OF RIVER FLOW DATA BY GENERALIZED PARETO DISTRIBUTION KUMER PIAL DAS Department of Mathematics,

More information

PERFORMANCE OF PARAMETER ESTIMATION TECHNIQUES WITH INHOMOGENEOUS DATASETS OF EXTREME WATER LEVELS ALONG THE DUTCH COAST.

PERFORMANCE OF PARAMETER ESTIMATION TECHNIQUES WITH INHOMOGENEOUS DATASETS OF EXTREME WATER LEVELS ALONG THE DUTCH COAST. PERFORMANCE OF PARAMETER ESTIMATION TECHNIQUES WITH INHOMOGENEOUS DATASETS OF EXTREME WATER LEVELS ALONG THE DUTCH COAST. P.H.A.J.M. VAN GELDER TU Delft, Faculty of Civil Engineering, Stevinweg 1, 2628CN

More information

ESTIMATING BIVARIATE TAIL

ESTIMATING BIVARIATE TAIL Elena DI BERNARDINO b joint work with Clémentine PRIEUR a and Véronique MAUME-DESCHAMPS b a LJK, Université Joseph Fourier, Grenoble 1 b Laboratoire SAF, ISFA, Université Lyon 1 Framework Goal: estimating

More information

Estimation of Quantiles

Estimation of Quantiles 9 Estimation of Quantiles The notion of quantiles was introduced in Section 3.2: recall that a quantile x α for an r.v. X is a constant such that P(X x α )=1 α. (9.1) In this chapter we examine quantiles

More information

Irr. Statistical Methods in Experimental Physics. 2nd Edition. Frederick James. World Scientific. CERN, Switzerland

Irr. Statistical Methods in Experimental Physics. 2nd Edition. Frederick James. World Scientific. CERN, Switzerland Frederick James CERN, Switzerland Statistical Methods in Experimental Physics 2nd Edition r i Irr 1- r ri Ibn World Scientific NEW JERSEY LONDON SINGAPORE BEIJING SHANGHAI HONG KONG TAIPEI CHENNAI CONTENTS

More information

Stat 5101 Lecture Notes

Stat 5101 Lecture Notes Stat 5101 Lecture Notes Charles J. Geyer Copyright 1998, 1999, 2000, 2001 by Charles J. Geyer May 7, 2001 ii Stat 5101 (Geyer) Course Notes Contents 1 Random Variables and Change of Variables 1 1.1 Random

More information

Variable inspection plans for continuous populations with unknown short tail distributions

Variable inspection plans for continuous populations with unknown short tail distributions Variable inspection plans for continuous populations with unknown short tail distributions Wolfgang Kössler Abstract The ordinary variable inspection plans are sensitive to deviations from the normality

More information

Models and estimation.

Models and estimation. Bivariate generalized Pareto distribution practice: Models and estimation. Eötvös Loránd University, Budapest, Hungary 7 June 2011, ASMDA Conference, Rome, Italy Problem How can we properly estimate the

More information

Investigation of an Automated Approach to Threshold Selection for Generalized Pareto

Investigation of an Automated Approach to Threshold Selection for Generalized Pareto Investigation of an Automated Approach to Threshold Selection for Generalized Pareto Kate R. Saunders Supervisors: Peter Taylor & David Karoly University of Melbourne April 8, 2015 Outline 1 Extreme Value

More information

Introduction. Chapter 1

Introduction. Chapter 1 Chapter 1 Introduction In this book we will be concerned with supervised learning, which is the problem of learning input-output mappings from empirical data (the training dataset). Depending on the characteristics

More information