Zum Hauptinhalt springen

Cookies bei Macro by Mark.Wir verwenden notwendige Cookies für Anmeldung, Sprache und Sicherheit. Mit Ihrer Einwilligung verwenden wir außerdem optionale Cookies für Präferenzen, für Analysen und für Werbemessung mit dem Meta-(Facebook-)Pixel. Lesen Sie unsere Cookie-Richtlinie.

Macro by MarkMacro by Mark
  • Start
Wird geladen …
Wird geladen …
← Modelle
MacroKonzepteWirtschaftspolitikModelle

SARIMA Model

Seasonal ARIMA - extends ARIMA with explicit seasonal AR, differencing, and MA terms for series with calendar structure.

ÜberblickDiagrammMathematikQuellen

Understand the mathematics

Follow the same equations and assumptions one section at a time or in a continuous reading.

Hover or focus a symbol defined in this model to find its exact matches. Select it to keep the highlight; Escape clears the selection.

Seasonal Model Specification and Notation

The SARIMA(p,d,q)(P,D,Q)s\text{SARIMA}(p,d,q)(P,D,Q)_sSARIMA(p,d,q)(P,D,Q)s​ model operates on a scalar time series yty_tyt​ observed at equally spaced dates with known seasonal period sss. Six orders govern the model: non-seasonal AR order ppp, ordinary differencing order ddd, non-seasonal MA order qqq, seasonal AR order PPP, seasonal differencing order DDD, and seasonal MA order QQQ. The model assumes that after ddd ordinary and DDD seasonal differences, the resulting series is a covariance-stationary process whose autocorrelation structure factors into a non-seasonal and a seasonal multiplicative component.

The multiplicative structure is the core distinction from non-seasonal ARIMA. Instead of fitting a single high-order ARMA polynomial to capture both within-season and between-season dependence, SARIMA factors the dependence into two low-order polynomials whose product generates the full lag structure. This factoring reduces the parameter count from O(p + Ps) free coefficients (in an unconstrained high-order ARMA) to p + q + P + Q parameters, while still generating interaction terms at lags like s+1 and s-1 that capture the cross-frequency dynamics.

ϕ(L) Φ(Ls) (1−L)d(1−Ls)D yt=c+θ(L) Θ(Ls) εt\phi(L)\,\Phi(L^s)\,(1-L)^d(1-L^s)^D\, y_t = c + \theta(L)\,\Theta(L^s)\,\varepsilon_tϕ(L)Φ(Ls)(1−L)d(1−Ls)Dyt​=c+θ(L)Θ(Ls)εt​
ϕ(L)\phi(L)ϕ(L)Φ(Ls)\Phi(L^s)Φ(Ls)(1−L)d(1−Ls)D(1-L)^d(1-L^s)^D(1−L)d(1−Ls)Dyty_tyt​cccθ(L)\theta(L)θ(L)Θ(Ls)\Theta(L^s)Θ(Ls)εt\varepsilon_tεt​

The compact SARIMA(p,d,q)(P,D,Q)s\text{SARIMA}(p,d,q)(P,D,Q)_sSARIMA(p,d,q)(P,D,Q)s​ representation. The AR polynomials multiply on the left; the MA polynomials multiply on the right; the two differencing operators sit between them.

ϕ(L)=1−ϕ1L−ϕ2L2−⋯−ϕpLp\phi(L) = 1 - \phi_1 L - \phi_2 L^2 - \cdots - \phi_p L^pϕ(L)=1−ϕ1​L−ϕ2​L2−⋯−ϕp​Lp
ϕ(L)\phi(L)ϕ(L)ϕ1\phi_1ϕ1​LLL

Non-seasonal AR polynomial of order p, with roots outside the unit circle for stationarity.

Φ(Ls)=1−Φ1Ls−Φ2L2s−⋯−ΦPLPs\Phi(L^s) = 1 - \Phi_1 L^s - \Phi_2 L^{2s} - \cdots - \Phi_P L^{Ps}Φ(Ls)=1−Φ1​Ls−Φ2​L2s−⋯−ΦP​LPs
Φ(Ls)\Phi(L^s)Φ(Ls)Φ1\Phi_1Φ1​sss

Seasonal AR polynomial of order P, operating at lags s, 2s, ..., Ps.

θ(L)=1+θ1L+θ2L2+⋯+θqLq\theta(L) = 1 + \theta_1 L + \theta_2 L^2 + \cdots + \theta_q L^qθ(L)=1+θ1​L+θ2​L2+⋯+θq​Lq
θ(L)\theta(L)θ(L)θ1\theta_1θ1​LLL

Non-seasonal MA polynomial of order q, with roots outside the unit circle for invertibility.

Θ(Ls)=1+Θ1Ls+Θ2L2s+⋯+ΘQLQs\Theta(L^s) = 1 + \Theta_1 L^s + \Theta_2 L^{2s} + \cdots + \Theta_Q L^{Qs}Θ(Ls)=1+Θ1​Ls+Θ2​L2s+⋯+ΘQ​LQs
Θ(Ls)\Theta(L^s)Θ(Ls)Θ1\Theta_1Θ1​sss

Seasonal MA polynomial of order Q, operating at seasonal lags.

ϕ(L) Φ(Ls)=1−ϕ1L−⋯−ϕpLp−Φ1Ls+ϕ1Φ1Ls+1+⋯\phi(L)\,\Phi(L^s) = 1 - \phi_1 L - \cdots - \phi_p L^p - \Phi_1 L^s + \phi_1\Phi_1 L^{s+1} + \cdotsϕ(L)Φ(Ls)=1−ϕ1​L−⋯−ϕp​Lp−Φ1​Ls+ϕ1​Φ1​Ls+1+⋯
ϕ(L)\phi(L)ϕ(L)Φ(Ls)\Phi(L^s)Φ(Ls)ϕ1\phi_1ϕ1​LLLΦ1\Phi_1Φ1​sss

Expanding the product AR polynomial reveals cross-terms at lags s+1, s+2, ..., s+p. These cross-terms are not free parameters - they are products of the non-seasonal and seasonal AR coefficients.

Notations

yty_tyt​

Observed series

wtw_twt​

Doubly-differenced series

LLL

Lag operator

sss

Seasonal period

ϕ(L)\phi(L)ϕ(L)

Non-seasonal AR polynomial

Φ(Ls)\Phi(L^s)Φ(Ls)

Seasonal AR polynomial

θ(L)\theta(L)θ(L)

Non-seasonal MA polynomial

Θ(Ls)\Theta(L^s)Θ(Ls)

Seasonal MA polynomial

εt\varepsilon_tεt​

Innovation

About these symbols

Observed series:

Doubly-differenced series: wt=(1−L)d(1−Ls)Dytw_t = (1-L)^d(1-L^s)^D y_twt​=(1−L)d(1−Ls)Dyt​

Lag operator: Lkyt=yt−kL^k y_t = y_{t-k}Lkyt​=yt−k​

Seasonal period: e.g. 12 for monthly, 4 for quarterly

Non-seasonal AR polynomial:

Seasonal AR polynomial:

Non-seasonal MA polynomial:

Seasonal MA polynomial:

Innovation: WN(0,σ2)\text{WN}(0, \sigma^2)WN(0,σ2)

    Transition checks

    Forecast origin
    y^T+h∣T=ET[yT+h],h=1,2,…\widehat{y}_{T+h|T}=\mathbb{E}_T[y_{T+h}],\qquad h=1,2,\ldotsy​T+h∣T​=ET​[yT+h​],h=1,2,…

    Conditioning on information at T separates the sample from the projection. The forecast can be smooth even when later realized data will not be.

    The bridge point is shared so the eye can see the handoff. After that point, the red line is a model forecast, not observed history.

    Do not score the forecast segment as realized data or read it as a policy promise.

    Uncertainty propagation
    Var⁡T(yT+h)=σ2∑j=0h−1ψj2\operatorname{Var}_T(y_{T+h})=\sigma^2\sum_{j=0}^{h-1}\psi_j^2VarT​(yT+h​)=σ2j=0∑h−1​ψj2​

    Each future innovation contributes through its impulse weight ψj\psi_jψj​. Unless the future is degenerate, longer horizons carry more conditional variance.

    The band widens when shocks accumulate through the model recursion.

    Do not treat the band edge as an alternative deterministic scenario.

    Variable timing

    Observed history

    yt∈ITy_t\in\mathcal{I}_Tyt​∈IT​
    yty_tyt​

    The history segment is realized data available at the forecast origin.

    Conditional mean

    y^T+h∣T=E[yT+h∣IT]\widehat{y}_{T+h|T}=\mathbb{E}[y_{T+h}\mid\mathcal{I}_T]y​T+h∣T​=E[yT+h​∣IT​]

    The forecast path is a conditional expectation, so it can be smooth while future realizations remain volatile.

    Prediction band

    σh2=Var⁡(yT+h∣IT)\sigma_h^2=\operatorname{Var}(y_{T+h}\mid\mathcal{I}_T)σh2​=Var(yT+h​∣IT​)

    Band width is horizon-specific uncertainty, not a second deterministic transition.

    Keep going

    Read next

    Continue with a related concept, model, or example.

    Theory model

    IS-LM Model

    a well-connected theoretical model two steps from here

    New angle: theoretical model →
    Macro by Mark
    Hilfe
    DatenschutzerklärungNutzungsbedingungenEthik und Compliance

    © 2026 Mark Jayson Nation LLC