|
Activity Number:
|
192
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 4, 2008 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Government Statistics
|
| Abstract - #302205 |
|
Title:
|
Model Utility in a Time Series with Interventions: A Case Study
|
|
Author(s):
|
William L. Seaver and Joy Oliver*+ and David J. Woehr and Brian K. Griepentrog
|
|
Companies:
|
The University of Tennessee, Knoxville and The University of Tennessee, Knoxville and Fors Marsh Group
|
|
Address:
|
408 Stokely Management Center, Knoxville, TN, 37996,
|
|
Keywords:
|
time series ; change points ; forecast horizons
|
|
Abstract:
|
Time series data with change points at the end of a series can make forecasting particularly complicated because these points can disrupt estimations of fit in the holdout sample. However, poor performance with a holdout sample is not a sign that the model is totally invalid. The current study sought to examine the predictive validity of various time series models, and then extended model fit estimations by forecasting with multiple horizons in order to address the complexity of the data. Implications of this study point to the importance of testing multiple origins and horizons for time series in which the effects of change points are difficult to unmask.
|
