I don't have a lot of background in TS analysis and I haven't found references that are clear on this. Hopefully someone here can help me understand this.

I'm optimising ARIMA coefficients for a multiple sets of orders. From my understanding, absolute value of MA coefficients should always be < 1 according due to the bound of invertibility. Then I've seen that same goes for AR coefficients (bound of stationary). Most of the references that I've found derive this with AR(p=1) or MA(q=1) but not for higher order models.

  • Would it be safe to set the constraint on all ARMA coefficients that they should be between -0.99 and 0.99.
  • Also, what would it mean if after the optimisation, the coefficients were pushed to the boundaries.


  • 1
    $\begingroup$ Does this answer your question? Regularization for ARIMA models $\endgroup$ – Adrian Keister Apr 2 at 22:24
  • $\begingroup$ "absolute value of AR/MA coefficients should always be < 1..." is not correct. $\endgroup$ – Michael Apr 3 at 14:38

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.