The University of Hong Kong · FACULTY OF STATISTICS

STAT8003 Chap.6 Yule-Walker Equations and AR Moments

- one subject, every graph, every model, every mark
6 Chapters4-page Bible
Our own words - no uploaded lecturer files
Updated for this semester
Chapter 6 of 8 · STAT8003

Yule-Walker Equations and AR Moments

Centre the process and derive the recurrence

This chapter derives the autocorrelation recursion for a stationary autoregressive process.

Multiplying the centred equation by a lagged process value and taking expectations removes the current innovation at positive lags, leaving the same recurrence that appears in the AR model.

Solve an initial block before moving forward

The first p correlations are linked through negative lags, which are replaced by symmetry. These Yule-Walker equations must be solved together with rho-zero equal to one.

After rho-one through rho-p are known, the recurrence supplies every later lag without a simultaneous system. Keeping that boundary visible prevents circular calculations.

Return to lag zero for variance

The current innovation survives when the equation is multiplied by the current process value. Its variance therefore appears in the gamma-zero equation.

Solve correlation first, then variance, and finally recover autocovariances as gamma-zero times rho-k. This order preserves the innovation term and the correct squared units.

In this chapter

What this chapter covers

  • 01

    Centred AR process

  • 02

    Autocorrelation recursion

  • 03

    Symmetry of correlations

  • 04

    Yule-Walker system

  • 05

    Lag-zero innovation term

  • 06

    AR variance

Worked example · free

Build the first AR(2) correlations

Q [9 marks]. AskSia-authored practice. The marks shown are an AskSia study allocation, not a University marking scheme. For phi1=0.3 and phi2=0.18, calculate rho1, rho2 and rho3.
The first correlations come from a simultaneous system.The first correlations come from a simultaneous system. Once rho one through rho p are known, the same autoregressive recursion generates every later lag.phi valuesand rho0=1solve initialrho1...rhoprecur forlater lagsYule-Walker workflow
The first correlations come from a simultaneous system. Once rho one through rho p are known, the same autoregressive recursion generates every later lag.
  • 3Use symmetry in the k=1 equation.
  • 3Calculate rho2 from rho1 and rho0.
  • 3Continue the recurrence to rho3.
The first equation is rho1=0.3+0.18rho1, so rho1 is approximately 0.36585. Then rho2=0.3rho1+0.18, approximately 0.28976, and rho3=0.3rho2+0.18rho1, approximately 0.15281.
Sia tip — Replace rho minus one by rho one before solving the first AR(2) equation.
Glossary

Key terms

Yule-Walker System
The simultaneous equations giving initial AR autocorrelations from model coefficients. They combine rho-zero equal to one with correlation symmetry to solve the first p lags.
ACF Recursion
The rule generating later autocorrelations from earlier ones and AR coefficients. Once the initial system is solved, the recurrence advances through all later lags without another simultaneous solve.
Innovation Variance
The variance of the new white-noise information entering each model period. It survives the lag-zero expectation and supplies the extra term needed to determine the process variance.
FAQ

Yule-Walker Equations and AR Moments FAQ

Why is symmetry needed in the first AR(2) equation?

At lag one, the recurrence includes rho minus one. Autocorrelation symmetry identifies this with rho one, turning the expression into an equation that can be solved for the unknown initial correlation.

Why does the variance derivation include an extra term?

At lag zero, the current innovation is part of Zt, so E(at Zt) equals the innovation variance. At positive lags, the current innovation is uncorrelated with the past process value and that term vanishes.

How does an AR ACF differ from an MA ACF?

The AR ACF generally follows a recurrence and tails away across infinitely many lags. The MA population ACF has finite support and cuts off after its order because values farther apart share no innovations.

In what order should AR moments be calculated?

Centre the process first, derive the positive-lag recurrence and solve the initial Yule-Walker system using rho-zero and symmetry. Continue later correlations with the forward recursion, keeping extra decimal places. Then return to lag zero, include the current innovation variance and solve gamma-zero. Multiply that variance by each correlation to recover autocovariances.

This order prevents the innovation term from disappearing and keeps units consistent.

What numerical checks should follow a Yule-Walker calculation?

Confirm rho-zero is one, every correlation magnitude is at most one and later values satisfy the recurrence. Keep unrounded initial correlations until the end. After computing gamma-zero, verify that the variance is positive and that recovered autocovariances have squared units.

Study strategy

Exam move

Derive the correlation equations before inserting numbers. Solve the initial Yule-Walker system with rho0=1 and symmetry, preserve several decimal places, and only then use the forward recurrence. Calculate variance last because it depends on the solved correlations. Check that all reported correlations remain between minus one and one.

Begin by centring the AR equation so constant terms do not obscure the covariance calculation. Multiply by the lagged process value, take expectations and state why the current innovation is uncorrelated with past values at positive lags. For the first p equations, replace negative correlations using symmetry and solve the resulting simultaneous system. Mark the boundary at p.

Beyond it, every right-hand correlation has already been found, so the recurrence moves forward without circularity. Recompute one later lag in two ways or substitute it back into the recursion as an arithmetic check. Keep unrounded values until the last displayed line because errors propagate along the tail. Then return to lag zero, where the current innovation no longer disappears.

Derive gamma zero with the innovation-variance term present and recover gamma k by multiplying the variance by rho k. Use units to distinguish the two stages: correlations are unitless, while autocovariances regain squared units. Compare the resulting tail with the hard cutoff of an MA model and explain why the patterns lead to different specification candidates.

If any correlation magnitude exceeds one or the computed variance is nonpositive, stop and trace the initial system before interpreting the shape.

Working through Yule-Walker Equations and AR Moments in STAT8003? Sia is AskSia’s AI Statistics tutor — ask any STAT8003 Yule-Walker Equations and AR Moments 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.

A+Everything unlocked
Unlocks this Bible + all 52 of your subjects at The University of Hong Kong, plus 1,000+ more Bibles across the AskSia library.
Sia - your STAT8003 tutor, unlimited, worked the way the exam marks it
The full 4-page Bible + practice bank with worked solutions
Chrome extension - sync your LMS so Sia knows your deadlines
Bilingual EN / Chinese on every Bible and every Sia answer
$0.99 Trial
30-day money-back · cancel in one tap · how it works