# Distribution of the Estimators for Autoregressive Time Series With a Unit Root

Опубликовано на портале: 06-04-2004
Journal of the American Statistical Association. 1979.  Vol. 74. No. 366. P. 427-431.
Тематический раздел:
Abstract Let $n$ observations $Y_1, Y_2, \ldots, Y_n$ be generated by the model $Y_t = \rho Y_{t - 1} + e_t$, where $Y_0$ is a fixed constant and $\{e_t\}_{t = 1}^n$ is a sequence of independent normal random variables with mean 0 and variance $\sigma^2$. Properties of the regression estimator of $\rho$ are obtained under the assumption that $\rho = \pm 1$. Representations for the limit distributions of the estimator of $\rho$ and of the regression $t$ test are derived. The estimator of $\rho$ and the regression $t$ test furnish methods of testing the hypothesis that $\rho = 1$.

David A. Dickey - Testing for Unit Roots Definition of a Unit Root Process and How to Test for Unit Roots

• AR(1) models
• Model: Yt - m = r ( Yt-1 - m ) + et
• Yt = observation at time t
• et = error or "shock" at time t (assumed iid normal)
• m = series mean (assumed constant over time)
• r = Autoregressive coefficient

• Test of Ho: r =1
• If r =1 then mean m drops out of model.
• If r =1 forecast does NOT revert to mean.
• Stock price example: if r =1 then can't make money by buying low and selling high.

• Test construction
• subtract ( Yt-1 - m ) from both sides of model
• Reparameterized Model: Yt - Yt-1 = (r -1)( Yt-1 - m ) + et
• Compute **First Difference** Dt = Yt-Yt-1
• Regress Dt on 1, Yt-1
• Test coefficient of Yt-1
• Distribution is nonstandard. n( r _hat - 1) is Op(1).
• Tables in Fuller, Introduction to Statistical Time Series
• Distribution (of coefficient)does NOT hold in higher order models (more lags)
• t-test on coefficient of Yt-1
• Distribution is nonstandard. t test is Op(1).
• Tables in Fuller, Introduction to Statistical Time Series
• Distribution of t DOES hold (is same asymptotically) in higher order models.

• Detrending

• Higher Order Models
• Yt - m = a 1 [Yt-1 - m ] + a 2 [Yt-2 - m ] +, ..., + a p[Yt-p -m ]+et

• Characteristic polynomial is:
• mp- a 1 mp-1 - a 2 mp-2 - ... - a p
• If m=1 is a root then 1 - a 1 - a 2 - ... - a p = 0 (definition of root)
• If m=1 then m [1 - a 1 - a 2 - ... - a p] is 0 regardless of m
• If m=1 then no mean m in model (not identifiable)
• No mean reversion
• Unit root is generalization of r =1 to higher order models.
• Compute first difference Dt = Yt -Yt-1 and its lagged values.
• Regress Dt= Yt -Yt-1 on 1, Yt-1, Dt-1 ,...,Dt-p-1
• True coefficient on Yt-1 is -[1 - a 1 - a 2 - ... - a p] (do the algebra!)
• Test coefficient of Yt-1 against t (t ) tables in Fuller.
• Test available in SAS (and other packages)
• Test discussed in Box and Jenkins (newest edition), Fuller, and many other texts.

BiBTeX
RIS
Ключевые слова

См. также:
Станислав О. Колеников
[Книга]
Halbert L. White
Econometrica. 1980.  Vol. 48. No. 4. P. 817-838.
[Статья]
Александр Олегович Крыштановский
Социология: методология, методы и математическое моделирование (Социология: 4М). 2000.  № 12. С. 96-112.
[Статья]
[Интернет-ресурс]
Randolph B. Cohen, Christopher Polk, Tuomo Vuolteenaho
Journal of Finance. 2003.  Vol. 58. No. 1. P. 609-641.
[Статья]
Jerry Hausman
Econometrica. 1978.  Vol. 46. No. 6. P. 1251-1272.
[Статья]