EViews.com

This thread is about the ARIMASel Add-in that perform automatic ARIMA selection, where the order of differencing is chosen via unit root tests, and the AR, SAR, MA and SMA terms are chosen according to an information criterion.

Change log

• 2010/02/10 Initial release
• 2010/04/12 Added error message for panel workfiles
• 2010/05/28 Changed seasonal terms so that they support a list of values to select between, not a simple maximum

There seems to be an error relating to KPSS and UROOT ("KPSS is an unrecognized option for UROOT."). It is in the Select Differencing section where it says:



Tchaithonov

I think I found out why - I ran it on a panel dataset. Maybe the code should ban the function from running? Just a thought.

Good point. I'll add one.

I have tried ARIMASel on a monthly series. Maybe I got something wrong, but I understand I need an SAR(12) to model a seasonal AR on monthly data. However, the ARIMASel menu only allows up to an SAR(3)...?

Unfortunately EViews won't let you add enough ARMA terms to make going up to 12 possible.

I suppose an alternative approach would be to let you specify which SAR/SMA terms you want to search between, rather than going all the way up to a maximum.

I'll contemplate it.

ok, just uploaded a new version that lets you select which seasonal terms to search within

EViews Gareth wrote:ok, just uploaded a new version that lets you select which seasonal terms to search within

Thank you Gareth, the add-in now seems to produce reasonable results on a monthly series with seasonality

Hello Gareth,
Thank you for add-in, it is really great.

I tried to use "dlog(aust).arimasel" in comment promth , but it gave error.
But when I try to use it without "dlog" in that case it works.

1) Is it possible to use dlog(series).arimasel?
I would like to use dlog of my series as independent variable

2) I dont want to use seasonal effect, If I just delete SAR SMA number from option menu, is it ok?

3) Can this add-in do AR MA optimization for not "sequenced ARMA" , I mean individual way: can it give an optimum value like that "AR(2) MA(2) MA(3)"together for a regresiion or just given in optimum value only in sequencenced ARMA like ARMA(2,3); AR(1) AR(2) MA(1),MA(2), MA(3) togetger.

Thanks

Create a new series that is equal to the dlog, then run the add-in on that series.

Just enter "0" as the seasonal terms.

Gareth,
Thank you for reply and your help.
It is really helpful program.

Is it possible to increase AR and MA to 20 instead of 10?
I know, you make reservation some ARMA parameters for SAR and SMA as there is limitation for usage of ARMA parameters in Eviews, but, in case I make 0 to SAR and SMA, Can I have a chance to increase AR and MA to higher number?
If yes, how?

Thanks

I don't think so.

Econometrically it is probably not a good idea anyway.

Hi, Gareth,

I am using ARIMASel to select the order of AR and MA. But everytime I change the maximum number of AR and MA, I will get different best AR and MA according to the AIC. Could you please tell me how I should specify the maximum number of AR and MA and how I can select the best AR and MA?

Thanks,

There isn't really an answer to that.

Thanks, Gareth,

I wonder if it is possible for you to take a look at the attached raw dataset and tell me how you think. I am new to Eview. I also want to ask you how to adjust the seasonality for this data series.