A Note on Second Order Conditions in Extreme Value Theory: Linking General and Heavy Tail Conditions (original) (raw)
Second order conditions ruling the rate of convergence in any first order condition involving regular variation and assuring a unified extreme value limiting distribution function for the sequence of maximum values, linearly normalized, have appeared in several contexts whenever researchers are working either with a general tail, i.e., γ ∈ R, or with heavy tails, with an extreme value index γ > 0. In this paper we shall clarify the link between the second order parameters, say ρ and ρ that have appeared in the two above mentioned set-ups, i.e., for a general tail and for heavy tails, respectively. We illustrate the theory with some examples and, for heavy tails, we provide a link with a third order framework.
Sign up for access to the world's latest research.
checkGet notified about relevant papers
checkSave papers to use in your research
checkJoin the discussion with peers
checkTrack your impact
Related papers
Laws of Small Numbers: Extremes and Rare Events, 2010
In this chapter we summarize results in extreme value theory, which are primarily based on the condition that the upper tail of the underlying df is in the δneighborhood of a generalized Pareto distribution (GPD). This condition, which looks a bit restrictive at first sight (see Section 2.2), is however essentially equivalent to the condition that rates of convergence in extreme value theory are at least of algebraic order (see Theorem 2.2.5). The δ-neighborhood is therefore a natural candidate to be considered, if one is interested in reasonable rates of convergence of the functional laws of small numbers in extreme value theory (Theorem 2.3.2) as well as of parameter estimators (Theorems 2.4.4, 2.4.5 and 2.5.4). 2.1 Domains of Attraction, von Mises Conditions Recall from Example 1.3.4 that a df F belongs to the domain of attraction of an extreme value df (EVD) G β (x) = exp(−(1 + βx) −1/β), 1 + βx > 0, denoted by F ∈ D(G β), iff there exist constants a n > 0, b n ∈ R such that F n (a n x + b n) −→ n→∞ G β (x), x∈ R ⇐⇒ P ((Z n:n − b n)/a n ≤ x) −→ n→∞ G β (x), x∈ R, where Z n:n is the sample maximum of an iid sample Z 1 ,. .. , Z n with common df F. Moreover, Z 1:n ≤ • • • ≤ Z n:n denote the pertaining order statistics. The Gnedenko-De Haan Theorem The following famous result due to Gnedenko [176] (and partially due to de Haan [184]) provides necessary as well as sufficient conditions for F ∈ D(G β).
On the estimation of the second order parameter in extreme-value theory
The extreme-value index γ plays an important parameter in extreme-value theory since it controls the first order behavior of the distribution tail. In the literature, numerous estimators of this parameter have been proposed especially in the case of heavy-tailed distributions, which is the situation considered here. Most of these estimators depend on the k largest observations of the underlying sample. Their bias is controlled by the second order parameter ρ. In order to reduce the bias of γ's estimators or to select the best number k of observations to use, the knowledge of ρ is essential. In this paper, we propose a simple approach to estimate the second order parameter ρ leading to both existing and new estimators. We establish a general result that can be used to easily prove the asymptotic normality of a large number of estimators proposed in the literature or to compare different estimators within a given family.
Rates of convergence in multivariate extreme value theory
Journal of Multivariate Analysis, 1991
We discuss rates of convergence for the distribution of normalized sample extremes to the appropriate limit distribution. We show that the rate of convergence depends on that of the corresponding dependence functions and that of the marginals. The univariate results are well known by now, so we restrict our attention to dependence functions (Sections 2 and 3). In the linal section of the paper we obtain a Berry-Es&en type result for multivariate extremes. 'i'l 1991 Academic Press, Inc.
On large deviation for extremes
Statistics & Probability Letters, 2003
Recently a weighted approximation for the tail empirical distribution function has been developed (Drees et al. ). We show that the same result can also be used to improve a known uniform approximation of the distribution of the maximum of a random sample. From this a general result about large deviations of this maximum is derived. In addition, the relationship between two second order conditions used in extreme value theory is clarified.
Weak Convergence (IIA) - Functional and Random Aspects of the Univariate Extreme Value Theory
arXiv: Probability, 2018
The univariate extreme value theory deals with the convergence in type of powers of elements of sequences of cumulative distribution functions on the real line when the power index gets infinite. In terms of convergence of random variables, this amounts to the the weak convergence, in the sense of probability measures weak convergence, of the partial maximas of a sequence of independent and identically distributed random variables. In this monograph, this theory is comprehensively studied in the broad frame of weak convergence of random vectors as exposed in Lo et al.(2016). It has two main parts. The first is devoted to its nice mathematical foundation. Most of the materials of this part is taken from the most essential Loeve(1936,177) and Haan (1970), based on the stunning theory of regular, pi or gamma variation. To prepare the statistical applications, a number contributions I made in my PhD and my Doctorate of Sciences are added in the last chapter of the last chapter of that p...
Loading Preview
Sorry, preview is currently unavailable. You can download the paper by clicking the button above.