I saw 2 types of formulas for ARIMA(0, 2, 2):

$$Ŷ_t = 2Y_{t-1} - Y_{t-2} + (α + β - 2)e_{t-1} + (1 - α)e_{t-2}$$ and $$y_t = -θ_1e_{t-1} - θ_2e_{t-2} + e_t \quad\quad (with \space y_t \space being \space 2nd \space difference \space of \space Y_t)$$

I'm confortable with the second formula which conform to the ARIMA(p,d,q) definition. is the first equivalent to second? what are α and β in the first formula? how is that the coeffient of $e_{t-1}$ term depends on both α and β?


They are equivalent models with different parameterizations.

The first form arises when showing that an ETS(A,A,N) model (a.k.a Holt's trend method) is equivalent to an ARIMA(0,2,2) model. The $\alpha$ and $\beta$ are the smoothing parameters from the ETS model. See https://otexts.com/fpp3/arima-ets.html

| cite | improve this answer | |
  • $\begingroup$ Thank you for the answer and comprehensive information $\endgroup$ – techie11 Jun 2 at 13:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.