Model order — ARIMA(p,d,q)

AR order p: 2 Differencing d: 1 MA order q: 1

Data generation

Trend drift: 0.030 %/mo Noise scale: 0.14

Forecast

Horizon: 24 months ahead

Stationarity check

Var(original y)—

Var(d-differenced)—

Variance ratio—

ADF t-stat (raw y)—

ADF t-stat (differenced)—

—

Fitted coefficients

φ: —

θ: —

Residual σ²—

AIC (approx.)—

Coefficients are estimated live from the generated data: the AR part via Yule-Walker/Levinson-Durbin on the differenced series, then AR+MA jointly refined with the Hannan-Rissanen conditional-least-squares method. Nothing here is hand-tuned — change p, d or q and the numbers above recompute from scratch.