Model selection and local geometry

DOI10.1214/19-AOS1940zbMATH Open1464.62304arXiv1801.08364OpenAlexW3110757481MaRDI QIDQ1996781FDOQ1996781

Authors: Yanyan Li 

Publication date: 26 February 2021

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

Abstract: We consider problems in model selection caused by the geometry of models close to their points of intersection. In some cases---including common classes of causal or graphical models, as well as time series models---distinct models may nevertheless have identical tangent spaces. This has two immediate consequences: first, in order to obtain constant power to reject one model in favour of another we need local alternative hypotheses that decrease to the null at a slower rate than the usual parametric $n^{-1/2}$ (typically we will require $n^{-1/4}$ or slower); in other words, to distinguish between the models we need large effect sizes or very large sample sizes. Second, we show that under even weaker conditions on their tangent cones, models in these classes cannot be made simultaneously convex by a reparameterization. This shows that Bayesian network models, amongst others, cannot be learned directly with a convex method similar to the graphical lasso. However, we are able to use our results to suggest methods for model selection that learn the tangent space directly, rather than the model itself. In particular, we give a generic algorithm for learning Bayesian network models.

Full work available at URL: https://arxiv.org/abs/1801.08364

Recommendations

• Counting probability distributions: Differential geometry and model selection
• Model selection for Gaussian concentration graphs
• A geometric view on learning Bayesian network structures
• SELECTING AMONGST LARGE CLASSES OF MODELS
• On the local geometry of mixture models

zbMATH Keywords

geometrymodel selectioncausal inferencehypothesis testasymptotic theoryconvex optimization

Mathematics Subject Classification ID

Parametric hypothesis testing (62F03) Hypothesis testing in multivariate analysis (62H15) Probabilistic graphical models (62H22) Statistics on metric spaces (62R20)

Cites Work

• Causation, prediction, and search
• Penalized estimation of directed acyclic graphs from discrete data
• Asymptotic Statistics
• Title not available (Why is that?)
• Statistics for high-dimensional data. Methods, theory and applications.
• PRMLT
• High-dimensional graphs and variable selection with the Lasso
• Title not available (Why is that?)
• Demystifying double robustness: a comparison of alternative strategies for estimating a population mean from incomplete data
• Adjusting for Nonignorable Drop-Out Using Semiparametric Nonresponse Models
• Pattern recognition and machine learning.
• Likelihood ratio tests and singularities
• Graphical models for associations between variables, some of which are qualitative and some quantitative
• Statistical modeling: The two cultures. (With comments and a rejoinder).
• Sparse inverse covariance estimation with the graphical lasso
• The log-linear group-lasso estimator and its asymptotic properties
• Inference on treatment effects after selection among high-dimensional controls
• Penalized likelihood methods for estimation of sparse high-dimensional directed acyclic graphs
• Causality. Models, reasoning, and inference
• Ancestral graph Markov models.
• Ideals, varieties, and algorithms. An introduction to computational algebraic geometry and commutative algebra
• Title not available (Why is that?)
• Lectures on algebraic statistics
• Learning Sparse Causal Gaussian Networks With Experimental Intervention: Regularization and Coordinate Descent
• Concave penalized estimation of sparse Gaussian Bayesian networks
• Geometry of the faithfulness assumption in causal inference
• Semiparametric theory for causal mediation analysis: efficiency bounds, multiple robustness and sensitivity analysis
• Differentiable manifolds.
• Marginal models for categorical data
• A linear non-Gaussian acyclic model for causal discovery
• Discrete chain graph models
• Marginal log-linear parameterization of conditional independence models
• Algebraic statistical models
• Role of models in statistical analysis
• On the completeness of orientation rules for causal discovery in the presence of latent confounders and selection bias
• Maximum likelihood estimation for linear Gaussian covariance models
• CAM: causal additive models, high-dimensional order search and penalized regression
• Title not available (Why is that?)
• Bayesian nonlinear model selection for gene regulatory networks
• Smoothness of marginal log-linear parameterizations

Cited In (5)

• Towards standard imsets for maximal ancestral graphs
• LOCALIZED MODEL SELECTION FOR REGRESSION
• Title not available (Why is that?)
• Selecting local models in multiple regression by maximizing power
• Model selection for simplicial approximation

Uses Software

• glasso
• TETRAD
• PRMLT

This page was built for publication: Model selection and local geometry

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1996781)