Advances and Applications in Statistics
Volume 50, Issue 6, Pages 435 - 467
(June 2017) http://dx.doi.org/10.17654/AS050060435 |
|
BOOTSTRAPPING AN AUTOREGRESSIVE TIME SERIES MODEL USING SAS
Maher Qumsiyeh, Robert Deis and Dalton Gannon
|
Abstract: In today’s world from science to finance being able to improve accuracy is a constant challenge. Statistical models such as the Box-Jenkins ARIMA model has been used to model and forecast data for years. However, many assumptions are made in these models that can affect the accuracy of their predictions when the assumptions are not met. The bootstrap method (introduced by Efron [9]) tries to circumvent these assumptions to find a better estimation of the true model. With this added accuracy, we are then able to model the data better and more accurately predict future data. In this paper, we will demonstrate how well the bootstrap performs in an auto-regressive model in comparison with the traditional Box-Jenkins methodology. Results were obtained using SAS on simulated data sets, with the error terms being Gaussian and non-Gaussian, as well as on a real data set. |
Keywords and phrases: bootstrap, time series, ARIMA, Box-Jenkins, simulation, SAS. |
|
Number of Downloads: 441 | Number of Views: 1166 |
|