Generalizations of the KPSS-test for stationarity

Bart Hobijn, Philip Hans Franses, Marius Ooms

Research output: Contribution to journalArticle

56 Citations (Scopus)

Abstract

We propose automatic generalizations of the KPSS-test for the null hypothesis of stationarity of a univariate time series. We can use these tests for the null hypotheses of trend stationarity, level stationarity and zero mean stationarity. We introduce the asymptotic null distributions and we determine consistency against relevant nonstationary alternatives. We compare the properties of the tests with those of other proposed tests for stationarity. Monte Carlo simulations support the relevance of the tests when an autoregressive process with large positive autocorrelations is likely under the null hypothesis.

Original languageEnglish (US)
Pages (from-to)483-502
Number of pages20
JournalStatistica Neerlandica
Volume58
Issue number4
DOIs
StatePublished - Nov 2004
Externally publishedYes

Fingerprint

Stationarity
Null hypothesis
Autoregressive Process
Null Distribution
Autocorrelation
Asymptotic distribution
Univariate
Time series
Monte Carlo Simulation
Likely
Generalization
KPSS test
Alternatives
Zero

Keywords

  • Bandwidth selection
  • Choi's test
  • Heteroskedasticity and autocorrelation consistent covariance estimation
  • Leybourne and McCabe's test
  • Long run variance
  • Rate of consistency
  • Stationarity test
  • Time series

ASJC Scopus subject areas

  • Statistics and Probability

Cite this

Generalizations of the KPSS-test for stationarity. / Hobijn, Bart; Franses, Philip Hans; Ooms, Marius.

In: Statistica Neerlandica, Vol. 58, No. 4, 11.2004, p. 483-502.

Research output: Contribution to journalArticle

Hobijn, Bart ; Franses, Philip Hans ; Ooms, Marius. / Generalizations of the KPSS-test for stationarity. In: Statistica Neerlandica. 2004 ; Vol. 58, No. 4. pp. 483-502.
@article{e4687a87667d41ef858c39785b953df5,
title = "Generalizations of the KPSS-test for stationarity",
abstract = "We propose automatic generalizations of the KPSS-test for the null hypothesis of stationarity of a univariate time series. We can use these tests for the null hypotheses of trend stationarity, level stationarity and zero mean stationarity. We introduce the asymptotic null distributions and we determine consistency against relevant nonstationary alternatives. We compare the properties of the tests with those of other proposed tests for stationarity. Monte Carlo simulations support the relevance of the tests when an autoregressive process with large positive autocorrelations is likely under the null hypothesis.",
keywords = "Bandwidth selection, Choi's test, Heteroskedasticity and autocorrelation consistent covariance estimation, Leybourne and McCabe's test, Long run variance, Rate of consistency, Stationarity test, Time series",
author = "Bart Hobijn and Franses, {Philip Hans} and Marius Ooms",
year = "2004",
month = "11",
doi = "10.1111/j.1467-9574.2004.00272.x",
language = "English (US)",
volume = "58",
pages = "483--502",
journal = "Statistica Neerlandica",
issn = "0039-0402",
publisher = "Wiley-Blackwell",
number = "4",

}

TY - JOUR

T1 - Generalizations of the KPSS-test for stationarity

AU - Hobijn, Bart

AU - Franses, Philip Hans

AU - Ooms, Marius

PY - 2004/11

Y1 - 2004/11

N2 - We propose automatic generalizations of the KPSS-test for the null hypothesis of stationarity of a univariate time series. We can use these tests for the null hypotheses of trend stationarity, level stationarity and zero mean stationarity. We introduce the asymptotic null distributions and we determine consistency against relevant nonstationary alternatives. We compare the properties of the tests with those of other proposed tests for stationarity. Monte Carlo simulations support the relevance of the tests when an autoregressive process with large positive autocorrelations is likely under the null hypothesis.

AB - We propose automatic generalizations of the KPSS-test for the null hypothesis of stationarity of a univariate time series. We can use these tests for the null hypotheses of trend stationarity, level stationarity and zero mean stationarity. We introduce the asymptotic null distributions and we determine consistency against relevant nonstationary alternatives. We compare the properties of the tests with those of other proposed tests for stationarity. Monte Carlo simulations support the relevance of the tests when an autoregressive process with large positive autocorrelations is likely under the null hypothesis.

KW - Bandwidth selection

KW - Choi's test

KW - Heteroskedasticity and autocorrelation consistent covariance estimation

KW - Leybourne and McCabe's test

KW - Long run variance

KW - Rate of consistency

KW - Stationarity test

KW - Time series

UR - http://www.scopus.com/inward/record.url?scp=10944241103&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=10944241103&partnerID=8YFLogxK

U2 - 10.1111/j.1467-9574.2004.00272.x

DO - 10.1111/j.1467-9574.2004.00272.x

M3 - Article

AN - SCOPUS:10944241103

VL - 58

SP - 483

EP - 502

JO - Statistica Neerlandica

JF - Statistica Neerlandica

SN - 0039-0402

IS - 4

ER -