Optimal regularized hypothesis testing in statistical inverse problems

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kretschmann, Remo, Wachsmuth, Daniel, Werner, Frank
Format: Preprint
Published: 2022
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913303051108352
author Kretschmann, Remo
Wachsmuth, Daniel
Werner, Frank
author_facet Kretschmann, Remo
Wachsmuth, Daniel
Werner, Frank
contents Testing of hypotheses is a well studied topic in mathematical statistics. Recently, this issue has also been addressed in the context of Inverse Problems, where the quantity of interest is not directly accessible but only after the inversion of a (potentially) ill-posed operator. In this study, we propose a regularized approach to hypothesis testing in Inverse Problems in the sense that the underlying estimators (or test statistics) are allowed to be biased. Under mild source-condition type assumptions we derive a family of tests with prescribed level $α$ and subsequently analyze how to choose the test with maximal power out of this family. As one major result we prove that regularized testing is always at least as good as (classical) unregularized testing. Furthermore, using tools from convex optimization, we provide an adaptive test by maximizing the power functional, which then outperforms previous unregularized tests in numerical simulations by several orders of magnitude.
format Preprint
id arxiv_https___arxiv_org_abs_2212_12897
institution arXiv
publishDate 2022
record_format arxiv
spellingShingle Optimal regularized hypothesis testing in statistical inverse problems
Kretschmann, Remo
Wachsmuth, Daniel
Werner, Frank
Statistics Theory
Numerical Analysis
62G10, 47A52, 65J20, 65F22, 65R30
Testing of hypotheses is a well studied topic in mathematical statistics. Recently, this issue has also been addressed in the context of Inverse Problems, where the quantity of interest is not directly accessible but only after the inversion of a (potentially) ill-posed operator. In this study, we propose a regularized approach to hypothesis testing in Inverse Problems in the sense that the underlying estimators (or test statistics) are allowed to be biased. Under mild source-condition type assumptions we derive a family of tests with prescribed level $α$ and subsequently analyze how to choose the test with maximal power out of this family. As one major result we prove that regularized testing is always at least as good as (classical) unregularized testing. Furthermore, using tools from convex optimization, we provide an adaptive test by maximizing the power functional, which then outperforms previous unregularized tests in numerical simulations by several orders of magnitude.
title Optimal regularized hypothesis testing in statistical inverse problems
topic Statistics Theory
Numerical Analysis
62G10, 47A52, 65J20, 65F22, 65R30
url https://arxiv.org/abs/2212.12897