Improving likelihood-ratio-based confidence intervals for threshold parameters in finite samples

Luiggi Donayre, Yunjong Eo, James Morley

Research output: Contribution to journalArticle

2 Citations (Scopus)

Abstract

Within the context of threshold regressions, we show that asymptotically-valid likelihood-ratio-based confidence intervals for threshold parameters perform poorly in finite samples when the threshold effect is large. A large threshold effect leads to a poor approximation of the profile likelihood in finite samples such that the conventional approach to constructing confidence intervals excludes the true threshold parameter value too often, resulting in low coverage rates. We propose a conservative modification to the standard likelihood-ratio-based confidence interval that has coverage rates at least as high as the nominal level, while still being informative in the sense of including relatively few observations of the threshold variable. An application to thresholds for US industrial production growth at a disaggregated level shows the empirical relevance of applying the proposed approach.

Original languageEnglish (US)
Article number20160084
JournalStudies in Nonlinear Dynamics and Econometrics
Volume22
Issue number1
DOIs
StatePublished - Feb 23 2018

Fingerprint

Threshold Parameter
Likelihood Ratio
Confidence interval
confidence
coverage
industrial production
Coverage
Profile Likelihood
regression
Categorical or nominal
Regression
Finite sample
Likelihood ratio
Valid
Approximation
Threshold effects

Keywords

  • confidence interval
  • finite-sample inference
  • inverted likelihood ratio
  • threshold regression

Cite this

Improving likelihood-ratio-based confidence intervals for threshold parameters in finite samples. / Donayre, Luiggi; Eo, Yunjong; Morley, James.

In: Studies in Nonlinear Dynamics and Econometrics, Vol. 22, No. 1, 20160084, 23.02.2018.

Research output: Contribution to journalArticle

@article{fcc1912e2dda40ce87f67d8fc9f5eefe,
title = "Improving likelihood-ratio-based confidence intervals for threshold parameters in finite samples",
abstract = "Within the context of threshold regressions, we show that asymptotically-valid likelihood-ratio-based confidence intervals for threshold parameters perform poorly in finite samples when the threshold effect is large. A large threshold effect leads to a poor approximation of the profile likelihood in finite samples such that the conventional approach to constructing confidence intervals excludes the true threshold parameter value too often, resulting in low coverage rates. We propose a conservative modification to the standard likelihood-ratio-based confidence interval that has coverage rates at least as high as the nominal level, while still being informative in the sense of including relatively few observations of the threshold variable. An application to thresholds for US industrial production growth at a disaggregated level shows the empirical relevance of applying the proposed approach.",
keywords = "confidence interval, finite-sample inference, inverted likelihood ratio, threshold regression",
author = "Luiggi Donayre and Yunjong Eo and James Morley",
year = "2018",
month = "2",
day = "23",
doi = "10.1515/snde-2016-0084",
language = "English (US)",
volume = "22",
journal = "Studies in Nonlinear Dynamics and Econometrics",
issn = "1081-1826",
publisher = "Berkeley Electronic Press",
number = "1",

}

TY - JOUR

T1 - Improving likelihood-ratio-based confidence intervals for threshold parameters in finite samples

AU - Donayre, Luiggi

AU - Eo, Yunjong

AU - Morley, James

PY - 2018/2/23

Y1 - 2018/2/23

N2 - Within the context of threshold regressions, we show that asymptotically-valid likelihood-ratio-based confidence intervals for threshold parameters perform poorly in finite samples when the threshold effect is large. A large threshold effect leads to a poor approximation of the profile likelihood in finite samples such that the conventional approach to constructing confidence intervals excludes the true threshold parameter value too often, resulting in low coverage rates. We propose a conservative modification to the standard likelihood-ratio-based confidence interval that has coverage rates at least as high as the nominal level, while still being informative in the sense of including relatively few observations of the threshold variable. An application to thresholds for US industrial production growth at a disaggregated level shows the empirical relevance of applying the proposed approach.

AB - Within the context of threshold regressions, we show that asymptotically-valid likelihood-ratio-based confidence intervals for threshold parameters perform poorly in finite samples when the threshold effect is large. A large threshold effect leads to a poor approximation of the profile likelihood in finite samples such that the conventional approach to constructing confidence intervals excludes the true threshold parameter value too often, resulting in low coverage rates. We propose a conservative modification to the standard likelihood-ratio-based confidence interval that has coverage rates at least as high as the nominal level, while still being informative in the sense of including relatively few observations of the threshold variable. An application to thresholds for US industrial production growth at a disaggregated level shows the empirical relevance of applying the proposed approach.

KW - confidence interval

KW - finite-sample inference

KW - inverted likelihood ratio

KW - threshold regression

UR - http://www.scopus.com/inward/record.url?scp=85037719470&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85037719470&partnerID=8YFLogxK

U2 - 10.1515/snde-2016-0084

DO - 10.1515/snde-2016-0084

M3 - Article

VL - 22

JO - Studies in Nonlinear Dynamics and Econometrics

JF - Studies in Nonlinear Dynamics and Econometrics

SN - 1081-1826

IS - 1

M1 - 20160084

ER -