Page 1 of 1

Automatic ARIMA selection - Too many ARMA terms

Posted: Wed Oct 02, 2019 9:29 am
by AfonsoRod
Dear all

I am trying to use the Automatic ARIMA selection add-in to select the correct ARMA terms to use. I read that, since I have daily data, I should test for 8 maximum number of AR and MA terms (7 days +1). Also, i also want to test my data for SAR and SMA terms. Hence, I specified ARMA add-in as follows:
fg.PNG
fg.PNG (7.23 KiB) Viewed 22136 times
However, after Eviews ends the estimation I get the following message:
Too many ARMA terms
I believe that I can only use 25 ARMA terms but, since I'm testing for a max of 8 AR terms, 8 MA terms, 2 SAR terms and 2 SMA terms, making a total of 20 max ARMA terms, shouldn't it be allowed? Or am I counting my ARMA terms the wrong way?

I appreciate all the help in advance

Yours sincerely,
Afonso Rodrigues

Re: Automatic ARIMA selection - Too many ARMA terms

Posted: Wed Oct 02, 2019 10:51 am
by EViews Gareth
Seasonal ARMA terms are multiplicative in the restriction.

Re: Automatic ARIMA selection - Too many ARMA terms

Posted: Wed Oct 02, 2019 11:14 am
by AfonsoRod
Seasonal ARMA terms are multiplicative in the restriction.
Eviews Gareth, thank you for your answer. But even if the seasonal ARMA terms are multiplicative shouldn't it still be possible to perform the test?

8 AR terms + 8 MA terms + ( 2 * 2 SAR terms) + ( 2 * 2 SMA terms ) = 24 ARMA terms

Also, even if I only opt to put in the SAR and SMA terms the option '0' and '7', I can only perform the test if I choose a max number of AR and MA terms of 5.

Re: Automatic ARIMA selection - Too many ARMA terms

Posted: Wed Oct 02, 2019 1:08 pm
by EViews Gareth
It is number of AR terms * number of SAR terms plus MA*SMA

Re: Automatic ARIMA selection - Too many ARMA terms

Posted: Wed Oct 02, 2019 1:54 pm
by AfonsoRod
It is number of AR terms * number of SAR terms plus MA*SMA
Thank you for the help

Re: Automatic ARIMA selection - Too many ARMA terms

Posted: Thu Oct 03, 2019 9:13 am
by AfonsoRod
It is number of AR terms * number of SAR terms plus MA*SMA
Eviews Gareth,

I'm sorry for the pressing, but in the tests I'm doing after your last help I tried to conduct two different tests, namely:
  • - 4 AR terms, 4 MA terms, 3 SAR terms and 3 SMA terms (0, 7 and 12)
  • - 6 AR terms, 6 MA terms, 2 SAR terms and 2 SMA terms (0, 7)
If I perform the equation you mentioned in your last comment we have, in both cases, a maximum number of 24 ARMA terms, since:
  • - 3 * 4 + 3 * 4 = 24
  • - 2 * 6 + 2 * 6 = 24
However, I can perform the test in the first case, but in the second case (with the 6 AR and MA terms and 2 SAR and SMA terms) the "too many ARMA terms" error pops up again.

What am i doing wrong?

Re: Automatic ARIMA selection - Too many ARMA terms

Posted: Thu Oct 03, 2019 9:21 am
by EViews Gareth
Sorry, I should have been more clear.

#AR + #MA + #AR*#SAR + #MA*#SMA

Re: Automatic ARIMA selection - Too many ARMA terms

Posted: Thu Oct 03, 2019 9:28 am
by AfonsoRod
Sorry, I should have been more clear.

#AR + #MA + #AR*#SAR + #MA*#SMA
Just one last thing: If in the "Searchable SAR terms" I write the options 0, 7 and 12, does that mean I have 3 SAR terms or just 2?

Re: Automatic ARIMA selection - Too many ARMA terms

Posted: Thu Oct 03, 2019 9:30 am
by EViews Gareth
Not sure.


As an aside, you should really use the built in automatic arima selection rather than the outdated add-in.

Re: Automatic ARIMA selection - Too many ARMA terms

Posted: Thu Oct 03, 2019 9:33 am
by AfonsoRod
Not sure.


As an aside, you should really use the built in automatic arima selection rather than the outdated add-in.
I do not have knowledge of that built in automatic arima selection. Where can I learn more about it?

Re: Automatic ARIMA selection - Too many ARMA terms

Posted: Thu Oct 03, 2019 9:34 am
by EViews Gareth

Re: Automatic ARIMA selection - Too many ARMA terms

Posted: Thu Oct 03, 2019 9:36 am
by AfonsoRod