Comparison of the bootstrap and bootstrap neural network methods in non linear time series

Document Type : Original Paper


1 Department of Statistics, University of Isfahan

2 Department of Information Technology Engineering , University of Isfahan


Neural networks are among those mathematical models which are used to model non-linear time series with high accuracy. The advantage with these linear times series as opposed to topical ones is that they don’t require restrictive assumptions. The accuracy of neural network based estimators as nonparametric models is of high importance. In that light, we can use bootstrapping to calculate the accuracy of estimators in the time series’ complex nonlinear structures. Though introduced in recent years these methods yield more accurate results in the bias calculation of estimators compared to the other ones. This paper introduces neural network bootstrap, bootstrap autoregressive, moving block bootstrap method and residual bootstrap methods in time series. Then these four algorithms are compared with each other in a simulation study. Finally an example related to Iran’s kerosene price monthly data is worked out.


Main Subjects

[1] Thombs, L.A. and Schucany, W.R. (1990). Bootstrap prediction intervals for autoregression, Journal of American Statistic Associat, 85, 486–492. [2] Alonso, A.M., Pena, D. and Romo, J. (2002). Forecasting time series with sieve bootstrap, Journal of
Statistical Planning and Inference, 100, 1–11. [3] Bühlmann, P. (1997). Sieve bootstrap for time series, Bernoulli, 3, 123– 148. [4] Bühlmann, P. (2002a). Bootstraps for time series, Statistical Science, 17, 52–72. [5] Giordano, F., La Rocca, M. and Perna, C.
(2007). Forecasting nonlinear time series with neural network sieve bootstrap. Computational Statistics and Data Analysis, 51, 3871–3884. [6] Bühlmann, P. (2002b). Sieve bootstrap with variable -length Markov Chains for stationary categorical time series, Journal of the American Statistical Association, 97, 443–471. [7] Giordano, F., La Rocca, M. and Perna, C. (2011). Properties of
the Neural network sieve bootstrap. Jorunal of Nonparametric Statistics, 23, 803-817. [8] Pérez-Rodríguez J.V. and Andrada-Félix J. (2013). Estimating critical values for testing the i.i.d in standardized residuals from GARCH models in finite sampeles. Computational Statistics, 28, 701–734. [9] Siani, C. and de Peretti, C. (2007). Analysing the performance of bootstrap neural tests for conditional heteroskedasticity in ARCH-M models, Computational Statistics and Data Analysis, 51, 2442–2460. [10]
Efron, B. (1979). Another look at the jackknife. Annals of Statistics. 7, 1–26. [11] Akaike, H. (1969). Fitting autoregressive models for prediction. Ann. Inst.Statist. Math. 21, 243–247. [12] Akaike, H. (1973). Information theory and an extension of the
maximum likelihood principle. Proceedings of the Second International Symposium on Information Theory. Akademiai Kiado, Budapest. [13] Efron, B. and Tibshirani, R. (1993). An introduction to the Bootstrap. Chapman and Hall.