Bayesian inference for multivariate extreme value distributions

Statistical modeling of multivariate and spatial extreme events has attracted broad attention in various areas of science. Max-stable distributions and processes are the natural class of models for this purpose, and many parametric families have been developed and successfully applied. Due to complicated likelihoods, the efficient statistical inference is still an active area of research, and usually composite likelihood methods based on bivariate densities only are used. Thibaud et al. (2016) use a Bayesian approach to fit a Brown–Resnick process to extreme temperatures. In this paper, we extend this idea to a methodology that is applicable to general max-stable distributions and that uses full likelihoods. We further provide simple conditions for the asymptotic normality of the median of the posterior distribution and verify them for the commonly used models in multivariate and spatial extreme value statistics. A simulation study shows that this point estimator is considerably more efficient than the composite likelihood estimator in a frequentist framework. From a Bayesian perspective, our approach opens the way for new techniques such as Bayesian model comparison in multivariate and spatial extremes. MSC 2010 subject classifications: Primary 62F15; secondary 60G70, 62F12.

[1]  M. Schlather,et al.  Estimation of Hüsler–Reiss distributions and Brown–Resnick processes , 2012, 1207.6886.

[2]  Marco Oesting,et al.  Asymptotic properties of the maximum likelihood estimator for multivariate extreme value distributions , 2016, 1612.05178.

[3]  Richard E. Chandler,et al.  Inference for clustered data using the independence loglikelihood , 2007 .

[4]  J. Segers,et al.  Multivariate peaks over thresholds models , 2016, 1603.06619.

[5]  Daoji Shi,et al.  Fisher information for a multivariate extreme value distribution , 1995 .

[6]  Sidney I. Resnick,et al.  Extreme values of independent stochastic processes , 1977 .

[7]  Anthony C. Davison,et al.  Bayesian inference for the Brown-Resnick process, with an application to extreme low temperatures , 2015, 1506.07836.

[8]  Jonathan A. Tawn,et al.  Exploiting occurrence times in likelihood inference for componentwise maxima , 2005 .

[9]  C. Zhou,et al.  On spatial extremes: With application to a rainfall problem , 2008, 0807.4092.

[10]  Stefano Castruccio,et al.  High-Order Composite Likelihood Inference for Max-Stable Distributions and Processes , 2014, 1411.0086.

[11]  A. Davison,et al.  Composite likelihood estimation for the Brown–Resnick process , 2013 .

[12]  Collin Carbno,et al.  Actuarial Theory for Dependent Risks: Measures, Orders, and Models , 2007, Technometrics.

[13]  J. L. Wadsworth,et al.  On the occurrence times of componentwise maxima and bias in likelihood inference for multivariate max-stable distributions , 2014, 1410.6733.

[14]  Raphael de Fondeville,et al.  High-dimensional peaks-over-threshold inference for the Brown-Resnick process , 2016, 1605.08558.

[15]  Thomas Opitz,et al.  Efficient inference and simulation for elliptical Pareto processes , 2013, 1401.0168.

[16]  P. Northrop,et al.  Posterior propriety in Bayesian extreme value analyses using reference priors , 2015, 1505.04983.

[17]  T. Opitz,et al.  Extremal tt processes: Elliptical domain of attraction and a spectral representation , 2012, J. Multivar. Anal..

[18]  Marc G. Genton,et al.  On the likelihood function of Gaussian max-stable processes , 2011 .

[19]  Anthony C. Davison,et al.  Extremes on river networks , 2015, 1501.02663.

[20]  Anthony C. Davison,et al.  A mixture model for multivariate extremes , 2007 .

[21]  Hector M. Ramos,et al.  A Sufficient Condition for Generalized Lorenz Order , 2000, J. Econ. Theory.

[22]  Laurens de Haan,et al.  Stationary max-stable fields associated to negative definite functions. , 2008, 0806.2780.

[23]  J. Tawn,et al.  Efficient inference for spatial extreme value processes associated to log-Gaussian random functions , 2014 .

[24]  S. Coles,et al.  Modelling Extreme Multivariate Events , 1991 .

[25]  Clément Dombry,et al.  Regular conditional distributions of continuous max-infinitely divisible random fields , 2013 .

[26]  Martin Schlather,et al.  Models for Stationary Max-Stable Random Fields , 2002 .

[27]  Aristidis K. Nikoloulopoulos,et al.  Extreme value properties of multivariate t copulas , 2009 .

[28]  Jonathan A. Tawn,et al.  Bayesian Inference for Extremes: Accounting for the Three Extremal Types , 2004 .

[29]  Anthony C. Davison,et al.  Geostatistics of extremes , 2012, Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences.

[30]  Marius Hofert,et al.  Nested Archimedean Copulas Meet R — The nacopula Package , 2011 .

[31]  J. Ivanovs,et al.  Robust bounds in multivariate extremes , 2016, 1608.04214.

[32]  J. Hüsler.  Maxima of normal random vectors: between independence and complete dependence , 1989 .

[33]  A. Davison,et al.  Bayesian Inference from Composite Likelihoods, with an Application to Spatial Extremes , 2009, 0911.5357.

[34]  S. Padoan,et al.  Likelihood-Based Inference for Max-Stable Processes , 2009, 0902.3060.

[35]  Marco Oesting,et al.  Exact simulation of max-stable processes. , 2015, Biometrika.

[36]  A. Davison,et al.  Likelihood estimators for multivariate extremes , 2014, 1411.3448.

[37]  Alexander Malinowski,et al.  Statistical Inference for Max-Stable Processes by Conditioning on Extreme Events , 2014, Advances in Applied Probability.

[38]  Mathieu Ribatet,et al.  Conditional simulation of max-stable processes , 2012, 1208.5376.

[39]  Johan Segers,et al.  An M‐estimator of spatial tail dependence , 2014, 1403.1975.

[40]  N. Tajvidi,et al.  Multivariate Generalized Pareto Distributions , 2006 .

[41]  C. Berg,et al.  Harmonic Analysis on Semigroups , 1984 .

[42]  Marc G. Genton,et al.  Full likelihood inference for max‐stable data , 2017, Stat.