Abstract
We deal with sequences of weakly dependent observations that are naturally ordered in time. Their constant mean is possibly subject to change at most once at some unknown time point. The aim is to test whether such an unknown change has occurred or not. The change point methods presented here rely on ratio type test statistics based on maxima of the cumulative sums. These detection procedures for the abrupt change in mean are also robustified by considering a general score function. The main advantage of the proposed approach is that the variance of the observations neither has to be known nor estimated. The asymptotic distribution of the test statistic under the no change null hypothesis is derived. Moreover, we prove the consistency of the test under the alternatives. A block bootstrap method is developed in order to obtain better approximations for the test’s critical values. The validity of the bootstrap algorithm is shown. The results are illustrated through a simulation study, which demonstrates computational efficiency of the procedures. A practical application to real data is presented as well.
Similar content being viewed by others
References
Andrews DWK (1991) Heteroskedasticity and autocorrelation consistent covariance matrix estimation. Econometrica 59(3):817–858
Andrews DWK (1993) Tests for parameter instability and structural change with unknown change point. Econometrica 61(4):821–856
Andrews DWK, Ploberger W (1994) Optimal tests when a nuisance parameter is present only under the alternative. Econometrica 62(6):1383–1414
Andrews DWK, Lee I, Ploberger W (1996) Optimal changepoint tests for normal linear regressions. J Econom 70(1):9–380
Antoch J, Hušková M (2001) Permutation tests in change point analysis. Stat Probab Lett 53(1):37–46
Antoch J, Hušková M, Prášková Z (1997) Effect of dependence on statistics for determination of change. J Stat Plan Infer 60(2):291–310
Bazarova A, Berkes I, Horváth L (2015) Change point detection with stable AR(1) errors. In: Dawson D, Kulik R, Haye MO, Szyszkowicz B, Zhao Y (eds) Asymptotic laws and methods in stochastics, fields institute communications, vol 76. Springer, Berlin, pp 179–193
Bradley RC (2005) Basic properties of strong mixing conditions. A survey and some open questions. Probab Surveys 2:107–144
Bulinskii AV (1987) Limit theorems under weak dependence conditions. In: Fourth international conference on probability theory and mathematical statistics. V.N.U. Sci Press, Utrecht, pp 307–326
Bulinskii AV (1989) On various conditions of mixing and asymptotic normality of random fields. Soviet Math Dokl 37(4):443–447
Chen Z, Tian Z (2014) Ratio tests for variance change in nonparametric regression. Statistics 48(1):1–16
Chen X, Wu Y (1989) Strong law for mixing sequence. Acta Math Appl Sin E 5(4):367–371
Csörgő M, Horváth L (1997) Limit theorems in change-point analysis. Wiley, Chichester
Doukhan P (1994) Mixing: properties and examples, Lecture Notes in Statistics, vol 85. Springer, New York
Fitzenberger B (1997) The moving block bootstrap and robust inference for linear least squares and quantile regression. J Econom 82:235–287
Hall P, Horowitz JL, Jing BY (1995) On blocking rules for the bootstrap with dependent data. Biometrika 82(3):561–574
Herrndorf N (1983) Stationary strongly mixing sequences not satisfying the central limit theorem. Ann Probab 11(3):809–813
Horváth L, Horváth Z, Hušková M (2008) Ratio tests for change point detection. In: Balakrishnan N, Peña EA, Silvapulle MJ (eds) Beyond parametrics in interdisciplinary research: Festschrift in Honor of Professor Pranab K. Sen, IMS Collections, Beachwood, Ohio, vol 1, pp 293–304
Hušková M (2004) Permutation principle and bootstrap in change point analysis. Fields Inst Commun 44:273–291
Hušková M (2007) Ratio type test statistics for detection of changes in time series. In: Gomes MI, Pinto Martins JA, Silva JA (eds) Bulletin of the International Statistical Institute, Proceedings of the 56th Session, Lisboa, vol 976, pp 3934–3937
Hušková M, Marušiaková M (2012) \(M\)-Procedures for detection of changes for dependent observations. Commun Stat B Simul 41(7):1032–1050
Ibragimov IA, Linnik YV (1971) Independent and stationary sequences of random variables. Wolters-Noordhoff, Groningen
Kim JY (2000) Detection of change in persistence of a linear time series. J Econom 95(1):97–116
Kim JY, Amador RB (2002) Corrigendum to detection of change in persistence of a linear time series. J Econom 109(2):389–392
Kirch C (2006) Resampling methods for the change analysis of dependent data. PhD thesis, University of Cologne, Germany
Lahiri S, Furukawa K, Lee YD (2007) A nonparametric plug-in rule for selecting optimal block lengths for block bootstrap methods. Stat Methodol 4(3):292–321
Kulperger RJ (1990) On the distribution of the maximum of brownian bridges with application to regression with correlated errors. J Stat Comput Sim 34(2–3):97–106
Künsch HR (1989) The jacknife and the bootstrap for general stationary observations. Ann Stat 17(3):1217–1241
Leybourne SJ, Taylor AR (2006) Persistence change tests and shifting stable autoregressions. Econ Lett 91(1):44–49
Madurkayová B (2011) Ratio type statistics for detection of changes in mean. Acta Univ Carolinae Math Phys 52(1):47–58
Perron P (2006) Dealing with structural breaks. In: Hassani H, Mills TC, Patterson K (eds) Palgrave handbook of econometrics, volume 1: econometric theory. Palgrave Macmillan, Basingstoke, pp 278–352
Peštová B, Pešta M (2015) Testing structural changes in panel data with small fixed panel size and bootstrap. Metrika 78(6):665–689
Peštová B, Pešta M (2016) Erratum to: Testing structural changes in panel data with small fixed panel size and bootstrap. Metrika 79(2):237–238
Politis DN, Romano JP (1992) A circular block-resampling procedure for stationary data. In: LePage R, Billard L (eds) Exploring the limits of bootstrap. Wiley, New York, pp 263–270
Politis DN, White H (2004) Automatic block-length selection for the dependent bootstrap. Econom Rev 23:53–70
Rosenblatt M (1971) Markov processes: structure and asymptotic behavior. Springer, Berlin
Wang D, Guo P, Xia Z (2016) Detection and estimation of structural change in heavy-tailed sequence. Commun Stat A Theor. https://doi.org/10.1080/03610926.1006780
Zhao W, Tian Z, Xia Z (2010) Ratio test for variance change point in linear process with long memory. Stat Pap 51:397–407
Zhao W, Xia Z, Tian Z (2011) Ratio test to detect change in the variance of linear process. Statistics 45:189–198
Acknowledgements
The authors would like to thank the Associate Editor and two anonymous referees for careful reading of the paper and for providing suggestions that improved this paper.
Author information
Authors and Affiliations
Corresponding author
Additional information
Institutional support to Barbora Peštová was provided by RVO:67985807. The research of Michal Pešta was supported by the Czech Science Foundation Project GAČR No. 15-04774Y.
Proofs
Proofs
Proof of Theorem 1
The proof is analogous in several steps with the proof of Theorem 1.1 in Horváth et al. (2008). Without loss on generality, we assume that \(\mu =0\). Let
By applying Theorem 1 from Doukhan (1994, Section 1.5.1) with the consequent remark justified by Herrndorf (1983) and Bulinskii (1987, 1989), we get
where \(\{\mathcal {W}(t),\,0\le t\le 1\}\) is a standard Wiener process and \(\widetilde{\mathcal {W}}(t)=\mathcal {W}(1)-\mathcal {W}(t)\). Consequently, Lemma 4.3 and Lemma 4.4 by Hušková and Marušiaková (2012) together with Assumptions A1–A5 lead to
for some \(\kappa >0\), where [a] denotes the integer part of \(a\in \mathbb {R}\). Hence,
Similarly, we get
With respect to (10), we get for all \(0<\gamma <1/2\)
The continuous mapping theorem completes the proof
\(\square \)
Proof of Theorem 2
Let \(k>\tau +1\) and \(k=[\xi n]\) for some \(\zeta<\xi <1-\gamma \). Note that \(\tau =O(n)\) and \(k=O(n)\) as \(n\rightarrow \infty \). By the mean value theorem, we get
where \(\mu ^*\) lies between \(\mu \) and \(\widehat{\mu }_{1k}(\psi )\). The first sum in (11) can be expanded using Lemma 4.3 by Hušková and Marušiaková (2012) and Assumptions A1–A4 as
as \(k\rightarrow \infty \) for any \(\theta \in [-1/2,0]\) and \(\nu \in \left( 0,\eta /(3(2+\chi +\chi '))\right) \). The Taylor expansion of \(\psi \) in the neighborhood of 0 with respect to Assumption A4 provides
The second sum in (11) can be rewritten using Lemma 4.4 by Hušková and Marušiaková (2012) with respect to the Lipschitz property from Assumption A4 as
as \(k\rightarrow \infty \) for any \(\theta \in [-1/2,0]\). Combining (11)–(14), we end up with
since \(\delta _n=O(k^{\theta })\) for some \(\theta \in \left( -\frac{1}{2},\frac{\eta }{3(2+\chi +\chi ')}-\frac{1}{2}\right) \subset [-1/2,0]\).
Consequently, applying Lemma 4.3 by Hušková and Marušiaková (2012) again, we obtain
Assumption A1 allows us to apply law of large numbers for \(\alpha \)-mixing sequences (Chen and Wu 1989). Hence,
uniformly for \(i=1,\ldots ,k\). Furthermore, due to the mean value theorem and Assumption A4, we have
where \(e^*\) is between 0 and \(\mu -\widehat{\mu }_{1k}(\psi )+\mathcal {I}\{j>\tau ,l>\tau \}\). Plugging (15) into (18) yields
Let us take into account (16) together with (17), (18), and (19). Thus, we obtain
as \(k\rightarrow \infty \). Let us choose \(\nu =\theta +1/2\). Thus, \(\theta \in \left( -\frac{1}{2},\frac{\eta }{3(2+\chi +\chi ')}-\frac{1}{2}\right) \) iff \(\nu \in \left( 0,\frac{\eta }{3(2+\chi +\chi ')}\right) \). Since \(\delta _n=O\left( k^{\theta }\right) \) as \(k\rightarrow \infty \) and
according to the proof of Theorem 1 (requiring Assumption A5), we get
Note that there is no change in the means of \(Y_k,\ldots ,Y_n\). Again from the proof of Theorem 1, we have
which completes the proof. \(\square \)
Proof of Theorem 3
Let us denote \(\mathsf {P}\,^*(\cdot )\equiv \mathsf {P}\,(\cdot |Y_1,\ldots ,Y_n)\). Moreover, let us define \(\beta _n=o_{\mathsf {P}\,,\mathsf {P}\,^*}(1), n\rightarrow \infty \) for some random sequence \(\{\beta _n\}_{n\in \mathbb {N}}\) as follows
Lemma 4.3 and Lemma 4.4 by Hušková and Marušiaková (2012) together with the mean value theorem used in a similar way as in the proof of Theorem 2 provide
Consequently, applying Lemma 4.3 by Hušková and Marušiaková (2012) again similarly as in the proof of Theorem 2, we obtain
Note that (20) contains \(S_{L,K}^{\varvec{U}}(p,q,l,k)\) and the expression in square brackets in (21)–(23) can be rewritten as
Furthermore, we define
If \(\{\varepsilon _i, i\in \mathbb {N}\}\) is an \(\alpha \)-mixing sequence, then \(\{\psi (\varepsilon _i), i\in \mathbb {N}\}\) is also \(\alpha \)-mixing, but with smaller or equal mixing coefficients than \(\{\varepsilon _i, i\in \mathbb {N}\}\) (Bradley 2005, Theorem 5.2). Proof of Theorem 3.6.2 by Kirch (2006) for \(q(t)=1,\,t\in (0,1)\); \(e(i)=\psi (\varepsilon _i)\); and \(d=-\lambda '(0)\delta _n\) together with Remarks 3.5.4 and 3.5.5 from Kirch (2006) provide, conditionally on \(\varepsilon _1,\ldots ,\varepsilon _n\),
in probability \(\mathsf {P}\,\) along \(\varepsilon _1,\ldots ,\varepsilon _n\). In contrast to Kirch (2006), we drop the assumption of random errors being a linear process (Kirch 2006, Remark 3.5.3), because this is only assumed in order to show that the original (not bootstrapped) statistic weakly converges under the null hypothesis. The assumptions of Theorem 3 and the null hypothesis (2) provide us the asymptotic distribution of \(\mathcal {R}_n(\psi ,\gamma )\) from Theorem 1.
By the uniform stochastic closeness (20)–(23) in a \(\mathsf {P}\,\)-stochastic sense and (24), we get, conditionally on \(Y_1,\ldots ,Y_n\),
in probability \(\mathsf {P}\,\) along \(Y_1,\ldots ,Y_n\). Finally, the assertion of Theorem 3 is straightforward, since the considered bootstrap statistic is a continuous function of the above vector of statistics. \(\square \)
Rights and permissions
About this article
Cite this article
Peštová, B., Pešta, M. Abrupt change in mean using block bootstrap and avoiding variance estimation. Comput Stat 33, 413–441 (2018). https://doi.org/10.1007/s00180-017-0785-4
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00180-017-0785-4