STAT8003 Chap.5 Autoregressive Models and Root Conditions
Autoregressive Models and Root Conditions
Do not confuse an equation with its stationary process
This chapter treats an autoregressive equation as a model that still needs a stationary solution. Past process values feed back into the current value, but that feedback is stable only under the characteristic-root condition.
Writing an AR label does not prove the condition.
Test every root before expanding
Form the polynomial with the course sign convention, solve its roots and compute each magnitude. Every root must lie strictly outside the unit circle. A boundary root fails. For AR(1), this becomes the familiar absolute-coefficient condition below one.
Complex roots use their distance from the origin, not only their real part.
Interpret the stationary innovation weights
After the root test passes, solve the mean from expectations and generate the psi weights. AR(1) weights decay geometrically, while AR(2) weights follow a two-term recursion. Psi-j measures how a shock from j periods ago enters the current value.
Substitution back into the original equation verifies both the intercept and each innovation coefficient.
What this chapter covers
- 01
AR model equation
- 02
Characteristic polynomial
- 03
Unit-circle root test
- 04
Stationary mean
- 05
Innovation expansion
- 06
Recursive psi weights
Worked example · free
Solve a stationary AR(1)
- 3Find and classify the characteristic root.
- 3Solve the expectation equation.
- 3Generate the innovation weights.
Key terms
- Characteristic Polynomial
- The lag polynomial whose roots determine the AR stationarity condition. Every root must have magnitude strictly greater than one before the causal stationary innovation expansion is used.
- Unit Circle
- The complex-number boundary of magnitude one used in the root test. Roots on the boundary fail, while roots outside support decay in the corresponding stationary AR process.
- Stationary Solution
- The unique stable process satisfying an AR equation when its root condition holds. It has a constant mean and a causal innovation representation whose weights follow the AR recursion.
Autoregressive Models and Root Conditions FAQ
What is the AR(1) stationarity condition?
For Zt=theta0+phi Zt-1+at, the characteristic root is 1/phi. Requiring that root outside the unit circle is equivalent to requiring the absolute value of phi to be strictly below one.
Why is the AR intercept not usually the mean?
Taking expectations gives mu=theta0+phi1 mu+...+phip mu. Therefore the mean divides the intercept by one minus the sum of AR coefficients, unless that sum happens to be zero.
How are AR(2) innovation weights obtained?
Start with psi0=1 and psi1=phi1. Then apply psi_k=phi1 psi_k-1+phi2 psi_k-2. The recursion supplies the stationary expansion even when no simple closed form is needed.
What should a complete AR stationarity answer contain?
Write the characteristic polynomial, solve every root and report each magnitude. State that all roots must lie strictly outside the unit circle, so equality does not pass. After a successful test, solve the stationary mean and generate the innovation weights. If the roots are complex, use their full magnitudes.
Finish by interpreting sign and decay separately: magnitude governs stability, while sign or angle shapes the shock response.
How can an innovation expansion be checked?
Substitute the proposed expansion for the current and lagged process values in the original AR equation. Match the intercept and the coefficients of the current shock and several lagged shocks. Equality in expectation alone is not enough to verify the representation.
Exam move
Always use the same order: write the polynomial, solve its roots, check every magnitude, find the stationary mean and then generate the psi weights. For negative roots or coefficients, distinguish magnitude from sign: magnitude decides the stationarity boundary, while sign helps describe oscillation or alternation in the response to a shock.
Create a root table with the polynomial, each real or complex root, its magnitude and the verdict. Treat equality with one as failure because the condition requires roots strictly outside. For AR(1), practise translating between the root statement and the equivalent absolute-coefficient condition until either direction is immediate.
For higher order models, keep the polynomial sign convention visible and verify roots by substitution when possible. After stationarity is established, solve the mean from the expectation equation rather than identifying it with the intercept. Generate psi weights recursively and interpret psi j as the effect of a shock from j periods ago.
Verify the first few weights by substituting the innovation expansion back into the original AR equation and matching coefficients. Compare a positive coefficient with a negative coefficient of the same magnitude: both may be stationary, but one produces same-direction decay while the other alternates. Finally, examine a boundary case and describe why the innovation effect does not decay adequately.
This routine turns the root test, stationary mean and shock response into one connected argument instead of three unrelated formulas. In a timed answer, show the polynomial and magnitude explicitly because the stationarity conclusion depends on them.
Working through Autoregressive Models and Root Conditions in STAT8003? Sia is AskSia’s AI Statistics tutor — ask any STAT8003 Autoregressive Models and Root Conditions question and get a clear, step-by-step explanation grounded in how STAT8003 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.