Maximum likelihood estimation for linear Gaussian covariance models

DOI10.1111/RSSB.12217MaRDI QIDQ5364909zbMATH OpenOpenAlexFDO

Authors Piotr Zwiernik, Caroline Uhler, Donald Richards

Publication date 29 September 2017

Published in Journal of the Royal Statistical Society Series B: Statistical Methodology (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1408.5604

zbMATH Keywords

Brownian motion tree modelconvex optimizationeigenvalues of random matriceslinear Gaussian covariance modelTracy-Widom lawWishart distribution

Mathematics Subject Classification ID

Point estimation (62F10) Parametric inference under constraints (62F30) Estimation in multivariate analysis (62H12)

Abstract: We study parameter estimation in linear Gaussian covariance models, which are $p$-dimensional Gaussian models with linear constraints on the covariance matrix. Maximum likelihood estimation for this class of models leads to a non-convex optimization problem which typically has many local maxima. Using recent results on the asymptotic distribution of extreme eigenvalues of the Wishart distribution, we provide sufficient conditions for any hill-climbing method to converge to the global maximum. Although we are primarily interested in the case in which $n>!!>p$, the proofs of our results utilize large-sample asymptotic theory under the scheme $n/pogamma>1$. Remarkably, our numerical simulations indicate that our results remain valid for $p$ as small as $2$. An important consequence of this analysis is that for sample sizes $nsimeq14p$, maximum likelihood estimation for linear Gaussian covariance models behaves as if it were a convex optimization problem.

Recommendations

• Maximum likelihood estimation of generalised linear models for multivariate normal covariance matrix
• Maximum likelihood estimation of parameter structures for the Wishart distribution using constraints
• Estimating linear covariance models with numerical nonlinear algebra
• Maximum likelihood mean and covariance matrix estimation constrained to general positive semi-definiteness
• Multivariate Gaussians, semidefinite matrix completion, and convex algebraic geometry

Cited in

(34)

• A note on the existence of maximum likelihood estimates for Gaussian-inverted Wishart models
• Model selection and local geometry
• Estimating variances in time series kriging using convex optimization and empirical BLUPs
• Hypothesis testing on linear structures of high-dimensional covariance matrix
• Maximum likelihood estimation in Gaussian models under total positivity
• Multivariate Gaussians, semidefinite matrix completion, and convex algebraic geometry
• Estimating linear covariance models with numerical nonlinear algebra
• Maximum likelihood degree of the two-dimensional linear Gaussian covariance model
• Covariance matrix estimation of the maximum likelihood estimator in multivariate clusterwise linear regression
• A survey of discrete methods in (algebraic) statistics for networks
• MAXIMUM LIKELIHOOD ESTIMATION IN LINEAR MODELS WITH EQUI-CORRELATED RANDOM ERRORS
• Maximum likelihood mean and covariance matrix estimation constrained to general positive semi-definiteness
• scientific article; zbMATH DE number 3984226 (Why is no real title available?)
• Maximum likelihood estimation of generalised linear models for multivariate normal covariance matrix
• A note on maximum likelihood estimation for covariance reducing models
• scientific article; zbMATH DE number 6982327 (Why is no real title available?)
• Covariance Structure Maximum-Likelihood Estimates in Compound Gaussian Noise: Existence and Algorithm Analysis
• scientific article; zbMATH DE number 910669 (Why is no real title available?)
• Likelihood geometry of correlation models
• Averaging estimation for conditional covariance models
• Brownian motion tree models are toric
• Nonparametric estimation of isotropic covariance function
• Modeling the Cholesky factors of covariance matrices of multivariate longitudinal data
• Likelihood landscape and maximum likelihood estimation for the discrete orbit recovery model
• MLE of jointly constrained mean-covariance of multivariate normal distributions
• Positive-definite thresholding estimators of covariance matrices with zeros
• Logarithmic Voronoi cells for Gaussian models
• Modelling correlation matrices in multivariate data, with application to reciprocity and complementarity of child-parent exchanges of support
• An integrated test on the linear structure of high-dimensional covariance matrices
• High-Dimensional Covariate-Dependent Gaussian Graphical Models
• Testing for patterns and structures in covariance and correlation matrices
• Entropic covariance models
• Linear covariance selection model via _1-penalization
• Maximum likelihood estimation for Brownian motion tree models based on one sample

This page was built for publication: Maximum likelihood estimation for linear Gaussian covariance models

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5364909)