Maximum likelihood estimation in Gaussian models under total positivity

DOI10.1214/17-AOS1668zbMATH Open1467.62092arXiv1702.04031WikidataQ127838243 ScholiaQ127838243MaRDI QIDQ2313271FDOQ2313271

Authors: Caroline Uhler, Piotr Zwiernik, Steffen Lauritzen 

Publication date: 18 July 2019

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

Abstract: We analyze the problem of maximum likelihood estimation for Gaussian distributions that are multivariate totally positive of order two (MTP2). By exploiting connections to phylogenetics and single-linkage clustering, we give a simple proof that the maximum likelihood estimator (MLE) for such distributions exists based on at least 2 observations, irrespective of the underlying dimension. Slawski and Hein, who first proved this result, also provided empirical evidence showing that the MTP2 constraint serves as an implicit regularizer and leads to sparsity in the estimated inverse covariance matrix, determining what we name the ML graph. We show that we can find an upper bound for the ML graph by adding edges corresponding to correlations in excess of those explained by the maximum weight spanning forest of the correlation matrix. Moreover, we provide globally convergent coordinate descent algorithms for calculating the MLE under the MTP2 constraint which are structurally similar to iterative proportional scaling. We conclude the paper with a discussion of signed MTP2 distributions.

Full work available at URL: https://arxiv.org/abs/1702.04031

Recommendations

• Maximum likelihood estimation for linear Gaussian covariance models
• scientific article; zbMATH DE number 775123
• Maximum likelihood estimation for Gaussian processes under inequality constraints
• Maximum Likelihood Estimation for a Smooth Gaussian Random Field Model
• Maximum likelihood estimation in linear infinite dimensional models
• Positive definite maximum likelihood covariance estimators
• scientific article; zbMATH DE number 3886910

zbMATH Keywords

Gaussian graphical modelultrametricinverse M-matrix\(\mathrm{MTP}_2\) distributionattractive Gaussian Markov random field (GMRF)nonfrustrated GRMF

Mathematics Subject Classification ID

Multivariate analysis (62H99) Estimation in multivariate analysis (62H12) Inequalities; stochastic orderings (60E15) Positive matrices and their generalizations; cones of matrices (15B48)

Cites Work

• Graphical models with R.
• Title not available (Why is that?)
• Title not available (Why is that?)
• Towards a Unified Theory of Inequality Constrained Testing in Multivariate Analysis
• Gaussian Markov distributions over finite graphs
• Classes of orderings of measures and related correlation inequalities. I. Multivariate totally positive distributions
• A likelihood ratio test for \(\text{MTP}_ 2\) within binary variables.
• Some notions of multivariate positive dependence
• Inverse \(M\)-matrices and ultrametric matrices
• Correlation inequalities on some partially ordered sets
• Exact sampling with coupled Markov chains and applications to statistical mechanics
• Total positivity in Markov structures
• The completion problem for \(M\)-matrices and inverse \(M\)-matrices
• Learning latent tree graphical models
• High-dimensional Gaussian graphical model selection: walk summability and local separation criterion
• Title not available (Why is that?)
• Geometry of maximum likelihood estimation in Gaussian graphical models
• Total positivity properties of absolute value multinormal variables with applications to confidence interval estimates and related probabilistic inequalities
• On the convergence of the coordinate descent method for convex differentiable minimization
• Path product matrices
• A recursive algorithm for Markov random fields
• Covariance selection and multivariate dependence
• Title not available (Why is that?)
• The correlation space of Gaussian latent tree models and model selection without fitting
• Über die Determinanten mit überwiegender Hauptdiagonale
• Inverse \(M\)-matrices. II
• Title not available (Why is that?)
• A general central limit theorem for FKG systems
• Title not available (Why is that?)
• M-matrices as covariance matrices of multinormal distributions
• Estimation of positive definite \(M\)-matrices and structure learning for attractive Gaussian Markov random fields
• The maximum likelihood threshold of a graph
• Learning Graphs With Monotone Topology Properties and Multiple Connected Components

Cited In (22)

• Total positivity in exponential families with application to binary variables
• Maximum likelihood estimation for totally positive log-concave densities
• Total positivity in multivariate extremes
• Maximum likelihood geometry in the presence of data zeros
• Dependence in elliptical partial correlation graphs
• High-Order Analysis of the Efficiency Gap for Maximum Likelihood Estimation in Nonlinear Gaussian Models
• Likelihood-based estimation for Gaussian MRFs
• Universal inference
• The geometry of gaussoids
• Nonlinear algebra and applications
• Maximum likelihood degree of the two-dimensional linear Gaussian covariance model
• On Sidorenko's conjecture for determinants and Gaussian Markov random fields
• Tropical Gaussians: a brief survey
• A multitest procedure for testing MTP2 for Gaussian distributions
• Learning graph Laplacian with MCP
• Convolutions of totally positive distributions with applications to kernel density estimation
• Bimonotone subdivisions of point configurations in the plane
• Total positivity in Markov structures
• Optimal rates for estimation of two-dimensional totally positive distributions
• Locally associated graphical models and mixed convex exponential families
• Gaussian Markov random fields and totally positive matrices
• Estimation of Monge matrices

Uses Software

• CVX

This page was built for publication: Maximum likelihood estimation in Gaussian models under total positivity

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2313271)