Total positivity in multivariate extremes

DOI10.1214/23-AOS2272arXiv2112.14727MaRDI QIDQ6136578FDOQ6136578

Sebastian Engelke, Frank Röttger, Piotr Zwiernik

Publication date: 31 August 2023

Published in: The Annals of Statistics (Search for Journal in Brave)

Abstract: Positive dependence is present in many real world data sets and has appealing stochastic properties that can be exploited in statistical modeling and in estimation. In particular, the notion of multivariate total positivity of order 2 ($mathrm{MTP}_2$) is a convex constraint and acts as an implicit regularizer in the Gaussian case. We study positive dependence in multivariate extremes and introduce $mathrm{EMTP}_2$, an extremal version of $mathrm{MTP}_2$. This notion turns out to appear prominently in extremes, and in fact, it is satisfied by many classical models. For a H"usler--Reiss distribution, the analogue of a Gaussian distribution in extremes, we show that it is $mathrm{EMTP}_2$ if and only if its precision matrix is a Laplacian of a connected graph. We propose an estimator for the parameters of the H"usler--Reiss distribution under $mathrm{EMTP}_2$ as the solution of a convex optimization problem with Laplacian constraint. We prove that this estimator is consistent and typically yields a sparse model with possibly nondecomposable extremal graphical structure. Applying our methods to a data set of Danube River flows, we illustrate this regularization and the superior performance compared to existing methods.

Full work available at URL: https://arxiv.org/abs/2112.14727

Recommendations

• scientific article; zbMATH DE number 3896082
• Extremes for multivariate expectiles
• Inequalities for the extremal coefficients of multivariate extreme value distributions
• Multivariate extremes over a random number of observations
• scientific article; zbMATH DE number 34422
• Statistics of Multivariate Extremes
• scientific article; zbMATH DE number 3885130
• scientific article

zbMATH Keywords

convex optimizationextreme value theorytotal positivitygraph Laplacianscovariance mapping

Mathematics Subject Classification ID

Extreme value theory; extremal stochastic processes (60G70) Probabilistic graphical models (62H22) Positive matrices and their generalizations; cones of matrices (15B48)

Cites Work

• OSQP: An Operator Splitting Solver for Quadratic Programs
• Graphical Models, Exponential Families, and Variational Inference
• Maxima of normal random vectors: Between independence and complete dependence
• Modelling multivariate extreme value distributions
• A Conditional Approach for Multivariate Extreme Values (with Discussion)
• Title not available (Why is that?)
• Maximum likelihood estimation for totally positive log‐concave densities
• Extreme value theory. An introduction.
• Title not available (Why is that?)
• Variograms for spatial max-stable random fields
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Causality. Models, reasoning, and inference
• Conditional association and unidimensionality in monotone latent variable models
• High-dimensional covariance estimation by minimizing \(\ell _{1}\)-penalized log-determinant divergence
• Extremes on river networks
• Title not available (Why is that?)
• Markov chain models for threshold exceedances
• A dependence measure for multivariate and spatial extreme values: Properties and inference
• Estimation of Hüsler–Reiss Distributions and Brown–Resnick Processes
• Exact simulation of max-stable processes
• An M-Estimator of Spatial Tail Dependence
• Association of Random Variables, with Applications
• Maximum empirical likelihood estimation of the spectral measure of an extreme-value distribu\-tion
• Dependence modelling for spatial extremes
• Dependence measures for extreme value analyses
• Interpolation of scattered data: distance matrices and conditionally positive definite functions
• Multivariate generalized Pareto distributions
• Properties of Euclidean and non-Euclidean distance matrices
• Classes of orderings of measures and related correlation inequalities. I. Multivariate totally positive distributions
• Some characterizations of symmetric inverse \(M\)-matrices
• A folded Laplace distribution
• Inverse \(M\)-matrices and ultrametric matrices
• Title not available (Why is that?)
• Conditional independence among max-stable laws
• Correlation inequalities on some partially ordered sets
• Geometry of cuts and metrics
• Simplicial matrix-tree theorems
• Exact tail asymptotics in bivariate scale mixture models
• Total positivity in Markov structures
• Recent Developments in Discrete Convex Analysis
• Positively correlated normal variables are associated
• Regular conditional distributions of continuous max-infinitely divisible random fields
• Limit theory for multivariate sample extremes
• Domains of attraction of multivariate extreme value distributions
• Phylogenetic trees and Euclidean embeddings
• Extremal dependence measure and extremogram: the regularly varying case
• Latent tree models
• A general central limit theorem for FKG systems
• Maximum likelihood estimation in Gaussian models under total positivity
• Estimation of positive definite \(M\)-matrices and structure learning for attractive Gaussian Markov random fields
• Graphical Models for Extremes
• Decompositions of dependence for high-dimensional extremes
• Total positivity in exponential families with application to binary variables
• Causal discovery in heavy-tailed models
• Factor Copula Models for Replicated Spatial Data
• Effective resistance is more than distance: Laplacians, simplices and the Schur complement
• Structure Learning for Extremal Tree Models
• Principal component analysis for multivariate extremes
• Locally associated graphical models and mixed convex exponential families
• Unifying Markov properties for graphical models
• Differentiating the pseudo determinant
• Extremal dependence of random scale constructions
• Dependence in elliptical partial correlation graphs
• One- versus multi-component regular variation and extremes of Markov trees
• Inference on extremal dependence in the domain of attraction of a structured Hüsler-Reiss distribution motivated by a Markov tree with latent variables
• Positive loadings and factor correlations from positive covariance matrices
• Optimization Algorithms for Graph Laplacian Estimation via ADMM and MM
• Brownian motion tree models are toric
• Spherical clustering in detection of groups of concomitant extremes
• Total positivity in multivariate extremes

Cited In (3)

• Total positivity in multivariate extremes
• Stochastic ordering in multivariate extremes
• Partial Tail-Correlation Coefficient Applied to Extremal-Network Learning

This page was built for publication: Total positivity in multivariate extremes

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6136578)