Optimal regularized hypothesis testing in statistical inverse problems

Abstract

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 $\alpha$ 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.

References in corpus (8)

• Frequentist coverage of adaptive nonparametric Bayesian credible sets
• Confidence bands in density estimation
• Risk hull method and regularization by projections of ill-posed inverse problems
• On adaptive inference and confidence bands
• Iteratively regularized Newton-type methods for general data misfit functionals and applications to Poisson data
• Minimax signal detection in ill-posed inverse problems
• Confidence bands for multivariate and time dependent inverse regression models
• The Le Cam distance between density estimation, Poisson processes and Gaussian white noise