Statistical Inference with Local Optima

DOI10.1080/01621459.2021.2023550MaRDI QIDQ6077585zbMATH OpenOpenAlexFDO

Authors

Publication date 18 October 2023

Published in Journal of the American Statistical Association (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1807.04431

zbMATH Keywords

maximum likelihood estimationnonconvexEM Algorithmtwo-sample testgradient descent

Mathematics Subject Classification ID

Statistics (62-XX)

Abstract: We study the statistical properties of an estimator derived by applying a gradient ascent method with multiple initializations to a multi-modal likelihood function. We derive the population quantity that is the target of this estimator and study the properties of confidence intervals (CIs) constructed from asymptotic normality and the bootstrap approach. In particular, we analyze the coverage deficiency due to finite number of random initializations. We also investigate the CIs by inverting the likelihood ratio test, the score test, and the Wald test, and we show that the resulting CIs may be very different. We propose a two-sample test procedure even when the MLE is intractable. In addition, we analyze the performance of the EM algorithm under random initializations and derive the coverage of a CI with a finite number of initializations.

Cites work

• scientific article; zbMATH DE number 3782216 (Why is no real title available?)
• scientific article; zbMATH DE number 3567782 (Why is no real title available?)
• scientific article; zbMATH DE number 3446222 (Why is no real title available?)
• scientific article; zbMATH DE number 2200453 (Why is no real title available?)
• scientific article; zbMATH DE number 3052144 (Why is no real title available?)
• A Statistical Learning Approach to Modal Regression
• A new regression model: modal linear regression
• All of Nonparametric Statistics
• Asymptotic Statistics
• Bootstrap methods: another look at the jackknife
• Gradient descent only converges to minimizers: non-isolated critical points and invariant regions
• Mixture Densities, Maximum Likelihood and the EM Algorithm
• Morse Theory. (AM-51)
• Note on the consistency of the maximum likelihood estimate for nonidentifiable distributions
• On Estimation of a Probability Density Function and Mode
• On the convergence properties of the EM algorithm
• On weak convergence and optimality of kernel density estimates of the mode
• Robust topological inference: distance to a measure and kernel distance
• Statistical analysis of finite mixture distributions
• Statistical guarantees for the EM algorithm: from population to sample-based analysis
• Statistical inference using the Morse-Smale complex
• Tests of Statistical Hypotheses Concerning Several Parameters When the Number of Observations is Large
• The landscape of empirical risk for nonconvex losses

Cited in

(2)

• Local statistical methods of knowledge formation
• Testing that a local optimum of the likelihood is globally optimum using reparameterized embeddings. Applications to wavefront sensing

This page was built for publication: Statistical Inference with Local Optima

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6077585)