Consistency of Bayes estimators of a binary regression function

DOI10.1214/009053606000000236MaRDI QIDQ2500452zbMATH OpenOpenAlexFDO

Publication date 24 August 2006

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

Full work available at URL https://arxiv.org/abs/math/0412203

Bayes procedure binary regression consistency entropy concavity large deviations subadditivity

Foundations and philosophical topics in statistics (62A01) Bayesian problems; characterization of Bayes procedures (62C10) Asymptotic distribution theory in statistics (62E20) Bayesian inference (62F15) Nonparametric inference (62G99)

Abstract: When do nonparametric Bayesian procedures ``overfit? To shed light on this question, we consider a binary regression problem in detail and establish frequentist consistency for a certain class of Bayes procedures based on hierarchical priors, called uniform mixture priors. These are defined as follows: let $u$ be any probability distribution on the nonnegative integers. To sample a function $f$ from the prior $p i^{u}$ , first sample $m$ from $u$ and then sample $f$ uniformly from the set of step functions from $[0, 1]$ into $[0, 1]$ that have exactly $m$ jumps (i.e., sample all $m$ jump locations and $m + 1$ function values independently and uniformly). The main result states that if a data-stream is generated according to any fixed, measurable binary-regression function $f_{0} o t e q u i v 1 / 2$ , then frequentist consistency obtains: that is, for any $u$ with infinite support, the posterior of $p i^{u}$ concentrates on any $L^{1}$ neighborhood of $f_{0}$ . Solution of an associated large-deviations problem is central to the consistency proof.

Recommendations

Cites work

Cited in

(9)

This page was built for publication: Consistency of Bayes estimators of a binary regression function

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2500452)