| 概要 |
In this article, we consider the asymptotic properties of time series models with a break in trend. We show that, for the model with a joint broken trend with stationary errors, the asymptotic propert...ies depend on the break magnitude. When the break magnitude is greater than 𝑂(𝑇^<−1/2>), the break date estimator is super-consistent, and the asymptotic properties of the estimators of the intercept and trend coefficients do not depend on whether the break date is known or estimated. These results indicate that the previously established results hold only under the “shrinking shift” asymptotic framework, in which the break magnitude shrinks to zero sufficiently fast. Simulation results illustrate that the finite sample approximation based on the proposed asymptotic theory works well.続きを見る
|
Mendeley出力
