|
Activity Number:
|
130
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Business and Economic Statistics Section
|
| Abstract - #304534 |
|
Title:
|
Sieve Bootstrap Prediction Intervals for Multivariate ARMA Models
|
|
Author(s):
|
Purna Mukhopadhyay*+ and V. A. Samaranayake
|
|
Companies:
|
The University of Kansas Medical Center and Missouri University of Science and Technology
|
|
Address:
|
Department of Biostatistics, Kansas City, KS, 66160,
|
|
Keywords:
|
Vector Time series ; Forecasting ; Bootstrapping ; Nonparametric
|
|
Abstract:
|
Currently available nonparametric bootstrap methods for obtaining prediction intervals for vector autoregressive moving average (ARMA) processes assume that the autoregressive and moving average orders, p, q respectively, are known. The sieve bootstrap method developed for univariate processes sidesteps the need for such an assumption for stationary and invertible time series. We adopt this univariate sieve bootstrap method for multivariate ARMA processes and further modify the technique to correct for residuals that underestimate the actual variance of the innovations and the failure of existing sieve bootstrap methods to capture variations due to sampling error of the mean. Monte Carlo simulations results show that the modified sieve bootstrap method provide prediction intervals that achieve nominal or near nominal coverage probabilities for bivariate ARMA processes.
|
