Inference for the mode of a log-concave density

Charles R. Doss; Jon A. Wellner

doi:10.1214/18-AOS1770

Inference for the mode of a log-concave density

Charles R. Doss, Jon A. Wellner

Statistics (Twin Cities)

Research output: Contribution to journal › Article › peer-review

8 Scopus citations

Abstract

We study a likelihood ratio test for the location of the mode of a logconcave density. Our test is based on comparison of the log-likelihoods corresponding to the unconstrained maximum likelihood estimator of a logconcave density and the constrained maximum likelihood estimator where the constraint is that the mode of the density is fixed, say at m. The constrained estimation problem is studied in detail in Doss and Wellner (2018). Here, the results of that paper are used to show that, under the null hypothesis (and strict curvature of -log f at the mode), the likelihood ratio statistic is asymptotically pivotal: that is, it converges in distribution to a limiting distribution which is free of nuisance parameters, thus playing the role of the χ₁ ² distribution in classical parametric statistical problems. By inverting this family of tests, we obtain new (likelihood ratio based) confidence intervals for the mode of a log-concave density f . These new intervals do not depend on any smoothing parameters. We study the new confidence intervals via Monte Carlo methods and illustrate them with two real data sets. The new intervals seem to have several advantages over existing procedures. Software implementing the test and confidence intervals is available in the R package logcondens.mode.

Original language	English (US)
Pages (from-to)	2950-2976
Number of pages	27
Journal	Annals of Statistics
Volume	47
Issue number	5
DOIs	https://doi.org/10.1214/18-AOS1770
State	Published - 2019

Bibliographical note

Funding Information:
Received December 2016; revised October 2018. 1Supported in part by NSF Grants DMS-1712664, DMS-1104832 and a University of Minnesota Grant-In-Aid grant. 2Supported in part by NSF Grants DMS-1104832 and DMS-1566514, NI-AID Grant 2R01 AI291968-04 and by the Isaac Newton Institute for Mathematical Sciences program Statistical Scalability, EPSRC Grant Number LNAG/036 RG91310. MSC2010 subject classifications. Primary 62G07; secondary 62G15, 62G10, 62G20. Key words and phrases. Mode, empirical processes, likelihood ratio, pivot, convex optimization, log-concave, shape constraints.

Funding Information:
Acknowledgments. We owe thanks to Lutz Dümbgen for several helpful conversations. This work was partially supported by a grant to the second author from the Simons Foundation and was carried out in part during a visit to the Isaac Newton Institute.

Funding Information:
Supported in part by NSF Grants DMS-1712664, DMS-1104832 and a University of Minnesota Grant-In-Aid grant. Supported in part by NSF Grants DMS-1104832 and DMS-1566514, NI-AID Grant 2R01 AI291968-04 and by the Isaac Newton Institute for Mathematical Sciences program Statistical Scalability, EPSRC Grant Number LNAG/036 RG91310

Publisher Copyright:
© Institute of Mathematical Statistics, 2019.

Keywords

Convex optimization
Empirical processes
Likelihood ratio
Log-concave
Mode
Pivot
Shape constraints

Access

10.1214/18-AOS1770

OpenUrl availability

Full text

Cite this

@article{c7a0743b75c849929b0c261eb3cec0b4,

title = "Inference for the mode of a log-concave density",

abstract = "We study a likelihood ratio test for the location of the mode of a logconcave density. Our test is based on comparison of the log-likelihoods corresponding to the unconstrained maximum likelihood estimator of a logconcave density and the constrained maximum likelihood estimator where the constraint is that the mode of the density is fixed, say at m. The constrained estimation problem is studied in detail in Doss and Wellner (2018). Here, the results of that paper are used to show that, under the null hypothesis (and strict curvature of -log f at the mode), the likelihood ratio statistic is asymptotically pivotal: that is, it converges in distribution to a limiting distribution which is free of nuisance parameters, thus playing the role of the χ1 2 distribution in classical parametric statistical problems. By inverting this family of tests, we obtain new (likelihood ratio based) confidence intervals for the mode of a log-concave density f . These new intervals do not depend on any smoothing parameters. We study the new confidence intervals via Monte Carlo methods and illustrate them with two real data sets. The new intervals seem to have several advantages over existing procedures. Software implementing the test and confidence intervals is available in the R package logcondens.mode.",

keywords = "Convex optimization, Empirical processes, Likelihood ratio, Log-concave, Mode, Pivot, Shape constraints",

author = "Doss, {Charles R.} and Wellner, {Jon A.}",

note = "Funding Information: Received December 2016; revised October 2018. 1Supported in part by NSF Grants DMS-1712664, DMS-1104832 and a University of Minnesota Grant-In-Aid grant. 2Supported in part by NSF Grants DMS-1104832 and DMS-1566514, NI-AID Grant 2R01 AI291968-04 and by the Isaac Newton Institute for Mathematical Sciences program Statistical Scalability, EPSRC Grant Number LNAG/036 RG91310. MSC2010 subject classifications. Primary 62G07; secondary 62G15, 62G10, 62G20. Key words and phrases. Mode, empirical processes, likelihood ratio, pivot, convex optimization, log-concave, shape constraints. Funding Information: Acknowledgments. We owe thanks to Lutz D{\"u}mbgen for several helpful conversations. This work was partially supported by a grant to the second author from the Simons Foundation and was carried out in part during a visit to the Isaac Newton Institute. Funding Information: Supported in part by NSF Grants DMS-1712664, DMS-1104832 and a University of Minnesota Grant-In-Aid grant. Supported in part by NSF Grants DMS-1104832 and DMS-1566514, NI-AID Grant 2R01 AI291968-04 and by the Isaac Newton Institute for Mathematical Sciences program Statistical Scalability, EPSRC Grant Number LNAG/036 RG91310 Publisher Copyright: {\textcopyright} Institute of Mathematical Statistics, 2019.",

year = "2019",

doi = "10.1214/18-AOS1770",

language = "English (US)",

volume = "47",

pages = "2950--2976",

journal = "Annals of Statistics",

issn = "0090-5364",

publisher = "Institute of Mathematical Statistics",

number = "5",

}

TY - JOUR

T1 - Inference for the mode of a log-concave density

AU - Doss, Charles R.

AU - Wellner, Jon A.

N1 - Funding Information: Received December 2016; revised October 2018. 1Supported in part by NSF Grants DMS-1712664, DMS-1104832 and a University of Minnesota Grant-In-Aid grant. 2Supported in part by NSF Grants DMS-1104832 and DMS-1566514, NI-AID Grant 2R01 AI291968-04 and by the Isaac Newton Institute for Mathematical Sciences program Statistical Scalability, EPSRC Grant Number LNAG/036 RG91310. MSC2010 subject classifications. Primary 62G07; secondary 62G15, 62G10, 62G20. Key words and phrases. Mode, empirical processes, likelihood ratio, pivot, convex optimization, log-concave, shape constraints. Funding Information: Acknowledgments. We owe thanks to Lutz Dümbgen for several helpful conversations. This work was partially supported by a grant to the second author from the Simons Foundation and was carried out in part during a visit to the Isaac Newton Institute. Funding Information: Supported in part by NSF Grants DMS-1712664, DMS-1104832 and a University of Minnesota Grant-In-Aid grant. Supported in part by NSF Grants DMS-1104832 and DMS-1566514, NI-AID Grant 2R01 AI291968-04 and by the Isaac Newton Institute for Mathematical Sciences program Statistical Scalability, EPSRC Grant Number LNAG/036 RG91310 Publisher Copyright: © Institute of Mathematical Statistics, 2019.

PY - 2019

Y1 - 2019

N2 - We study a likelihood ratio test for the location of the mode of a logconcave density. Our test is based on comparison of the log-likelihoods corresponding to the unconstrained maximum likelihood estimator of a logconcave density and the constrained maximum likelihood estimator where the constraint is that the mode of the density is fixed, say at m. The constrained estimation problem is studied in detail in Doss and Wellner (2018). Here, the results of that paper are used to show that, under the null hypothesis (and strict curvature of -log f at the mode), the likelihood ratio statistic is asymptotically pivotal: that is, it converges in distribution to a limiting distribution which is free of nuisance parameters, thus playing the role of the χ1 2 distribution in classical parametric statistical problems. By inverting this family of tests, we obtain new (likelihood ratio based) confidence intervals for the mode of a log-concave density f . These new intervals do not depend on any smoothing parameters. We study the new confidence intervals via Monte Carlo methods and illustrate them with two real data sets. The new intervals seem to have several advantages over existing procedures. Software implementing the test and confidence intervals is available in the R package logcondens.mode.

AB - We study a likelihood ratio test for the location of the mode of a logconcave density. Our test is based on comparison of the log-likelihoods corresponding to the unconstrained maximum likelihood estimator of a logconcave density and the constrained maximum likelihood estimator where the constraint is that the mode of the density is fixed, say at m. The constrained estimation problem is studied in detail in Doss and Wellner (2018). Here, the results of that paper are used to show that, under the null hypothesis (and strict curvature of -log f at the mode), the likelihood ratio statistic is asymptotically pivotal: that is, it converges in distribution to a limiting distribution which is free of nuisance parameters, thus playing the role of the χ1 2 distribution in classical parametric statistical problems. By inverting this family of tests, we obtain new (likelihood ratio based) confidence intervals for the mode of a log-concave density f . These new intervals do not depend on any smoothing parameters. We study the new confidence intervals via Monte Carlo methods and illustrate them with two real data sets. The new intervals seem to have several advantages over existing procedures. Software implementing the test and confidence intervals is available in the R package logcondens.mode.

KW - Convex optimization

KW - Empirical processes

KW - Likelihood ratio

KW - Log-concave

KW - Mode

KW - Pivot

KW - Shape constraints

UR - http://www.scopus.com/inward/record.url?scp=85071029080&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85071029080&partnerID=8YFLogxK

U2 - 10.1214/18-AOS1770

DO - 10.1214/18-AOS1770

M3 - Article

AN - SCOPUS:85071029080

SN - 0090-5364

VL - 47

SP - 2950

EP - 2976

JO - Annals of Statistics

JF - Annals of Statistics

IS - 5

ER -

Inference for the mode of a log-concave density

Abstract

Bibliographical note

Keywords

Access

OpenUrl availability

Other files and links

Fingerprint

Cite this