For time series that are not stationary, the block bootstrap method is not directly applicable. However, if the underlying stochastic structure is slowly changing with time, one may employ a local block-resampling procedure. We define such a procedure, and give an example of its applicability.
Pour les séries chronologiques qui ne sont pas stationnaires, la méthode de bloc re-échantillonnage n'est pas directement applicable. Cependant, si la structure stochastique fondamentale change lentement, on peut utiliser une méthode de bloc re-échantillonnage local. Nous définissons une telle procédure et donnons un exemple de son applicabilité.
Revised:
Published online:
Efstathios Paparoditis 1; Dimitris N. Politis 2
@article{CRMATH_2002__335_11_959_0, author = {Efstathios Paparoditis and Dimitris N. Politis}, title = {Local block bootstrap}, journal = {Comptes Rendus. Math\'ematique}, pages = {959--962}, publisher = {Elsevier}, volume = {335}, number = {11}, year = {2002}, doi = {10.1016/S1631-073X(02)02578-5}, language = {en}, }
Efstathios Paparoditis; Dimitris N. Politis. Local block bootstrap. Comptes Rendus. Mathématique, Volume 335 (2002) no. 11, pp. 959-962. doi : 10.1016/S1631-073X(02)02578-5. https://comptes-rendus.academie-sciences.fr/mathematique/articles/10.1016/S1631-073X(02)02578-5/
[1] On the Kullback–Leibler information divergence of locally stationary processes, Stochastic Process. Appl, Volume 62 (1996), pp. 139-168
[2] Fitting time series models to nonstationary processes, Ann. Statist, Volume 25 (1997), pp. 1-37
[3] The Jackknife and the bootstrap for general stationary observations, Ann. Statist, Volume 17 (1989), pp. 1217-1241
[4] Subsampling, Springer, New York, 1999
[5] Non-Linear and Non-Stationary Time Series Analysis, Academic Press, London, 1988
[6] Fixed design regression for time series: asymptotic normality, J. Multivariate Anal, Volume 40 (1992), pp. 262-291
[7] Local bootstrap, Ann. Inst. Statist. Math, Volume 43 (1991), pp. 667-676
Cited by Sources:
Comments - Policy