|
Activity Number:
|
15
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Sunday, August 2, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Business and Economic Statistics Section
|
| Abstract - #303705 |
|
Title:
|
Testing the Linearity Hypothesis in Nonlinear Autoregression
|
|
Author(s):
|
Michael Levine*+
|
|
Companies:
|
Purdue University
|
|
Address:
|
250 N. University St., Department of Statistics, West Lafayette, IN, 47907,
|
|
Keywords:
|
nonparametric ; autoregression ; ARCH ; model ; selection
|
|
Abstract:
|
Historically, model specification for the nonparametric time series models has been a difficult problem. This is particularly true if the conditional variance of the process is allowed to vary with time which is almost always true for financial time series. We consider a nonlinear ARCH model and propose a novel approach to checking the commonly used additive conditional mean structure. Our approach is based on testing the joint significance of all two-way interactive components. The proposed test is related to the unbalanced design ANOVA test with unequal variances. The "tuning" parameter selection is also addressed in detail. Simulation studies show that the method performs very well for sample sizes of about 5000 which are easily available in financial applications.
|
