An ARIMA model is specified by 3 parameters $(p,q,d)$ or 6 (+1 for the seasonality) if we consider a seasonal ARIMA model $(p,q,d)(P,Q,D)_s$.

The AIC used to select ARIMA models is calculated by:

$AIC = N log(\frac{SSE}{N}) + 2(k+2)$

With N the number of data points, SSE the sum of the squared errors, and K the number of parameters.

How is k calculated?

Is is simply $k=3$ for a regular ARIMA model and $k=6$ for a seasonal model?

Or is k to the number of coefficients of the polynomial: For example for an ARIMA$(3,0,2)$ model, $k = 5$ ?

If it is the second case, then doesn't differencing (meaning $q,Q \neq 0$) throw off the AIC completely, since that would change the value of N?


Your Answer

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

Browse other questions tagged or ask your own question.