Detecting a conditional extreme value model

In classical extreme value theory probabilities of extreme events are estimated assuming all the components of a random vector to be in a domain of attraction of an extreme value distribution. In contrast, the conditional extreme value model assumes a domain of attraction condition on a sub-collection of the components of a multivariate random vector. This model has been studied in Heffernan and Tawn (JRSS B 66(3):497–546, 2004), Heffernan and Resnick (Ann Appl Probab 17(2):537–571, 2007), and Das and Resnick (2009). In this paper we propose three statistics which act as tools to detect this model in a bivariate set-up. In addition, the proposed statistics also help to distinguish between two forms of the limit measure that is obtained in the model.     

Functional nonparametric estimation of conditional extreme quantiles

We address the estimation of quantiles from heavy-tailed distributions when functional covariate information is available and in the case where the order of the quantile converges to one as the sample size increases. Such “extreme” quantiles can be located in the range of the data or near and even beyond the boundary of the sample, depending on the convergence rate of their order to one. Nonparametric estimators of these functional extreme quantiles are introduced, their asymptotic distributions are established and their finite sample behavior is investigated.     

Estimation of the extreme value and the extreme points

Summary Letf be a continuous function defined on some domainA andX ₁,X ₂, ... be iid random variables. We estimate the extreme value off onA by studying the limiting distribution of min {f(X ₁), ...,f(X _n )} or max {f(X ₁), ...,f(X _n )} properly normalized. Sufficient conditions for the existence of the limiting distribution as well as a characterization of the limiting distribution relative to the extreme points off will be provided. A discussion of the multidimensional case is also carried out. Partially supported by CNPq-No. 301508/84.     

Transition kernels and the conditional extreme value model

Classical multivariate extreme value modelling assumes that the joint distribution belongs to a multivariate domain of attraction and this assumption requires that each marginal distribution be individually attracted to a univariate extreme value distribution. The Heffernan and Tawn (J R Stat Soc Ser B (Stat Methodol) 66(3):497–546, 2004) alternative extremal model for multivariate data does not require all the components belong to an extremal domain of attraction but assumes instead the existence of an asymptotic approximation to the conditional distribution of the random vector given one of the components is extreme. Combined with the knowledge that the conditioning component belongs to a univariate domain of attraction, this leads to an approximation of the probability of certain risk regions. The original focus on conditional distributions has technical drawbacks but is a natural assumption in many contexts. The technical drawbacks are overcome by relying on convergence of measures and the theory of extended regular variation Heffernan and Resnick (Ann Appl Probab 17(2):537–71, 2007); Das and Resnick (Extremes 14(1):29–61, 2000a); Das et al. (Adv Appl Probab 45(1):139–163, 2013). We compare the two approaches and describe in what way relying on variational limit properties of conditional distributions restricts the class of limit approximations.     

Estimation of a mean in the presence of an extreme observation

The performance of Anscombe, semi-Winsorization and Winsorization (A, S and W) rules for dealing with extreme observations are investigated for observations from N(μ, σ²) and the simple case where it is assumed that at most one observation in the sample may be biased, arising from N(μ + aσ, σ²) and the primary objective is to estimate μ when σ is unknown. Each of these rules is separately treated in terms of the estimated standard deviation, range and interquartile range. A Monte Carlo method is used to evaluate certain expectation integrals that arise in the computations. We give the results for sample sizes n = 6, 8, 10, 12, 14, 16, 20, 30, 40, 50, 60, 80, 100 of determining the constants necessary to give ‘premiums’ of 0.01 and 0.05 for each of the rules. The performance of the rules is measured in terms of ‘protection’. Features of the resulting tables are discussed.     

Conditions based on conditional moments for max-stable limit laws

Let X ₁, X ₂, ...X _n be independent and identically distributed random variables with common distribution function F. Necessary and sufficient conditions for F to belong to the domains of attraction of Φ _α and Ψ _α are derived in terms of conditional moments.      

On a multidimensional martingale with given conditional covariance structure

In this paper, a construction of an m-dimensional martingale with given conditional covariance structure is proposed. Application is given to certain stability problem of convergence rate in the martingale central limit theorem. Supported by the Lithuanian National Science Foundation and by the International Science Foundation (grant LI000). Proceedings of the XVI Seminar on Stability Problems for Stochastic Models, Part I, Eger, Hungary, 1994.     

On compact convex sets with given extreme points

The research was financially supported by the Russian Foundation for Fundamental Research (Grant 94-01-01286).     

On conditional extreme values of random vectors with polar representation

An effective approach for studying the asymptotics of bivariate random vectors is to search for the limits of conditional probabilities where the conditioning variable becomes large. In this context, elliptical and related distributions have been extensively investigated. A quite general model was presented by Fougères and Soulier (Limit conditional distributions for bivariate vectors with polar representation in Stochastic Models, 2010), who derived a conditional limit theorem for random vectors (X, Y) with a polar representation R · (u(T), v(T)), where R, T are stochastically independent and R is in the Gumbel max-domain of attraction. We reformulate their assumptions, such that they have a simpler structure, display more clearly the geometry of the curves (u(t), v(t)) and allow us to deduce interesting generalizations into two directions:

u has several global maxima instead of only one,

the curve (u(t), v(t)) is no longer differentiable, but forms a “cusp”.

The latter generalization yields results where only random norming leads to a non-degenerate limit statement. Ideas and results are elucidated by several figures.     

A conditional limit theorem for a bivariate representation of a univariate random variable and conditional extreme values

We first consider a real random variable X represented through a random pair (R,T) and a deterministic function u as X = R?u(T). Under quite weak assumptions we prove a limit theorem for (R,T) given X>x, as x tends to infinity. The novelty of our paper is to show that this theorem for the representation of the univariate random variable X permits us to obtain in an elegant manner conditional limit theorems for random pairs (X,Y) = R?(u(T),v(T)) given that X is large. Our approach allows to deduce new results as well as to recover under considerably weaker assumptions results obtained previously in the literature. Consequently, it provides a better understanding and systematization of limit statements for the conditional extreme value models.     

Estimation of the extreme value index and extreme quantiles under random censoring

In this paper, we consider the estimation of the extreme value index and extreme quantiles in the presence of random right censoring. The generalization of the peaks over threshold method is discussed and an adaptation of the moment estimator is proposed. The corresponding extreme quantile estimators are also introduced. We make a start with the analysis of the asymptotic properties of the moment estimator and the corresponding extreme quantile estimator. The finite sample behaviour is illustrated with a small simulation study and through practical examples from survival data analysis.      

Geometric stable laws: Estimation and applications

Geometric stable laws constitute a class of limiting distributions of appropriately normalized random sums of i.i.d. random variables. We consider the problem of estimation of the parameters of univariate and multivariate geometric stable laws. Our estimation technique is based on the method of moments and yields consistent and asymptotically normal estimators. We apply our estimators to a currency exchange data and show that the geometric stable dominates Paretian stable and normal models.     

Strong and conditional invariance principles for samples attracted to stable laws

Summary. We prove almost sure convergence of a representation of normalized partial sum processes of a sequence of i.i.d. random variables from the domain of attraction of an α-stable law, α<2. We obtain an explicit form of the limit in terms of the LePage series representation of stable laws. One consequence of these results is a conditional invariance principle having applications to option pricing as well as to resampling by signs and permutations. Received: 11 April 1994 / In revised form: 5 November 1996     

Estimation of the conditional stability of a two-contour boundary-value problem

We obtain estimates for the conditional stability of a two-contour boundary-value problem for fourth-order elliptic partial differential equations.Translated from Ukrainskii Matematicheskii Zhurnal, Vol. 47, No. 10, pp. 1418–1420, October, 1995.     

On the extreme points of the polytope of symmetric matrices with given row sums

The extreme points of the convex polytope of nonnegative symmetric matrices of order n with prescribed row sums are fully characterized by their respective graphs. For infinite matrices such a characterization is shown to be impossible. However, after imposing some additional conditions on the positive entries of the matrices, a considerable subfamily of infinite matrices is characterized by its graphs.