1
$\begingroup$

Here is the documentation for statsmodel's ARMA fit function. My issue is that it does not specify the form of the model it is fitting, i.e. is it fitting $a(L)y_t = b(L)\epsilon_t$ or is it fitting the regression form enter image description here

It's a bit confusing because statsmodel's ARMA generation assumes the form $a(L)y_t = b(L)\epsilon_t$, but the "no constant" option makes it unclear.

$\endgroup$
2
  • $\begingroup$ Since the documentation says ARMA(p,q), I think it has the form that you describe. It has the option to include or exclude the constant c. $\endgroup$ Jan 12, 2017 at 0:40
  • $\begingroup$ I mean, both are ARMA(p,q), one is just more succinct than the other. $\endgroup$
    – user369210
    Jan 12, 2017 at 0:41

1 Answer 1

0
$\begingroup$

The "constant" option in the fit returns the mean, from which the actual constant can be recovered using $\mu = \beta_0/(1- \beta_1 - \cdots - \beta_p)$.

$\endgroup$
1
  • $\begingroup$ Is this behavior documented? I couldn't find it on the package's website and spent days trying to figure out what the heck is going on. Thanks $\endgroup$
    – Confounded
    Apr 3, 2020 at 13:18

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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