A unit root is a feature of a time-series process that helps explain why its behavior can change over time and why ordinary statistical methods may not apply directly. This lesson introduces the definition of a unit root and shows how it makes a series non-stationary.
yt=a1yt−1+a2yt−2+⋯+apyt−p+εt.
The characteristic equation describes the process's roots. A process has a unit root when 1 is a solution to that equation. For the AR(p) form above, the equation is mp−a1mp−1−a2mp−2−⋯−ap=0. If m=1 is a root with multiplicity one, the process is integrated of order one, written I(1). If the root 1 has multiplicity r, the process is integrated of order r, written I(r).
All flashcards
What is a unit root?
A root equal to 1 in the characteristic equation of a stochastic process.
What does I(r) indicate?
A process with a unit root of multiplicity r; generally, r differences are needed to make it stationary.
When does an AR(1) model have a unit root?
For yt=a1yt−1+εt, when a1=1.
How is a random walk written?
As yt=yt−1+εt, where the current value is the previous value plus an innovation.
How do shocks behave in a unit-root process?
A shock can have a permanent effect on the series' level.
Quiz questions
What can happen when ordinary least squares is applied directly to non-stationary data?Answer: It can produce apparently strong but misleading regression results.
How do temporary shocks affect a trend-stationary series?Answer: They affect deviations from the trend, which tend to fade.
What does a characteristic root greater than 1 indicate?Answer: An explosive process.
What role does the drift term c play in yt=yt−1+c+et?Answer: It adds a systematic average change over time.
Based on the Wikipedia article Unit root, licensed CC BY-SA 4.0. Adapted into a summary, flashcards and quizzes.