Maximum likelihood estimation for totally positive log-concave densities

DOI10.1111/SJOS.12462zbMATH Open1473.62048arXiv1806.10120OpenAlexW3020812139MaRDI QIDQ5152170FDOQ5152170

Authors: Elina Robeva, Bernd Sturmfels, Caroline Uhler, Ngoc Mai Tran 

Publication date: 17 September 2021

Published in: Scandinavian Journal of Statistics (Search for Journal in Brave)

Abstract: We study nonparametric maximum likelihood estimation for two classes of multivariate distributions that imply strong forms of positive dependence; namely log-supermodular (MTP${}_2$) distributions and log-$$-concave (LLC) distributions. In both cases we also assume log-concavity in order to ensure boundedness of the likelihood function. Given $n$ independent and identically distributed random vectors in $mathbb{R}^d$ from one of our distributions, the maximum likelihood estimator (MLE) exists a.s. and is unique a.e. with probability one when $ngeq3$. This holds independently of the ambient dimension $d$. We conjecture that the MLE is always the exponential of a tent function. We prove this result for samples in ${0,1}^d$ or in $mathbb{R}^2$ under MTP${}_2$, and for samples in $mathbb{Q}^d$ under LLC. Finally, we provide a conditional gradient algorithm for computing the maximum likelihood estimate.

Full work available at URL: https://arxiv.org/abs/1806.10120

Recommendations

• Maximum likelihood estimation in Gaussian models under total positivity
• Maximum Likelihood Estimation of a Multi-Dimensional Log-Concave Density
• Theoretical properties of the log-concave maximum likelihood estimator of a multidimensional density
• Estimating a discrete log-concave distribution in higher dimensions
• Nonparametric estimation of multivariate convex-transformed densities

zbMATH Keywords

nonparametric density estimationlog-concavitytotal positivitysupermodularityshape-constrained density estimation

Mathematics Subject Classification ID

Density estimation (62G07) Characterization and structure theory of statistical distributions (62E10)

Cited In (10)

• Smoothed log-concave maximum likelihood estimation with applications
• Maximum likelihood estimation via duality for cell probabilities subject to convex and log-convex constraints
• Total positivity in multivariate extremes
• Dependence in elliptical partial correlation graphs
• Maximum likelihood estimation of the mixture of log-concave densities
• A new technique for calculating fourier integrals based on the poisson summation formula
• Universal inference
• Maximum-likelihood estimation of a log-concave density based on censored data
• Convolutions of totally positive distributions with applications to kernel density estimation
• Optimal rates for estimation of two-dimensional totally positive distributions

Uses Software

• LogConcDEAD

This page was built for publication: Maximum likelihood estimation for totally positive log-concave densities

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5152170)