|
Activity Number:
|
275
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, July 31, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
IMS
|
| Abstract - #309364 |
|
Title:
|
A Universal Nonparametric Test for Detecting Changes in Trend
|
|
Author(s):
|
Chuan Goh*+
|
|
Companies:
|
University of Toronto
|
|
Address:
|
Department of Economics, Toronto, ON, M5S 3G7, Canada
|
|
Keywords:
|
structural change ; large deviations ; limits of experiments ; partially linear model
|
|
Abstract:
|
A nonparametric test is proposed for detecting changes in the deterministic component of a dynamic time series model. The test is based on the asymptotic behavior of a quasi-likelihood ratio statistic similar to the class of procedures proposed by Fan et al (2001) for assessing the fit of nonparametric regression models in an iid-error setting. This test does not involve any asymptotic bias associated with parameter estimates of the stochastic component of the model when the deterministic trend component exhibits structural change. As such, the proposed test avoids a potentially important source of nonmonotonic power. Using the theory of large deviations, the procedure is also shown to be asymptotically optimal in a generalized Neyman-Pearson sense against global departures from the null of parameter stability when Gaussianity of the innovations terms is assumed.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2007 program |
