Page 2 of 6

Re: ARIMASel (Automatic ARIMA selection)

Posted: Mon Oct 18, 2010 7:36 pm
by jag_211085
hi,

how do i use this add-in? I can't seem to load it.

Re: ARIMASel (Automatic ARIMA selection)

Posted: Mon Oct 18, 2010 7:56 pm
by EViews Gareth
Do you have EViews 7.1?

Re: ARIMASel (Automatic ARIMA selection)

Posted: Mon Oct 18, 2010 8:12 pm
by jag_211085
No I have Eviews 6.

Is there any way to run it on Eviews 6?

Re: ARIMASel (Automatic ARIMA selection)

Posted: Mon Oct 18, 2010 9:09 pm
by EViews Gareth
No.

Re: ARIMASel (Automatic ARIMA selection)

Posted: Wed Jan 19, 2011 11:50 am
by ratcliff
Is there a way to capture ARIMASEL's recommended AR/MA specification within a program? I'd like to run arimasel on ~30 series, then create an equation object with the recommended specs. I'm currently working on accessing the crit table to find the minimum, but that seems a bit clunky. Is there temp variable, etc. (something like @nobs) that I can access with a program?

Thanks

Re: ARIMASel (Automatic ARIMA selection)

Posted: Wed Jan 19, 2011 11:55 am
by EViews Gareth
If you check the add-in program you can see that right at the end the best specification is stored in a program variable called %beststr. You could modify the program so that this is stored in a string object in your workfile. i.e. add something like this to the bottom of the program:

Code: Select all

string beststr = %beststr


You could then access that string from another program and/or the command line etc...


Edit:

Actually that might not do what you want, since the %beststr contains some other text. You could do this though:

Code: Select all

string chosenstr = {%eqname}.@spec


And put that line in at the bottom, prior to the line that says

Code: Select all

d {%eqname}

Re: ARIMASel (Automatic ARIMA selection)

Posted: Wed Jan 19, 2011 12:35 pm
by ratcliff
EViews Gareth wrote:Actually that might not do what you want, since the %beststr contains some other text. You could do this though:

Code: Select all

string chosenstr = {%eqname}.@spec


And put that line in at the bottom, prior to the line that says

Code: Select all

d {%eqname}


That worked great -- thanks!

Re: ARIMASel (Automatic ARIMA selection)

Posted: Tue May 03, 2011 6:17 am
by Sielenna
string chosenstr = {%eqname}.@spec
How can I change the above code such that Eviews gives automatically the name of the open series to the recommended specification?

Kind greetings,

Re: ARIMASel (Automatic ARIMA selection)

Posted: Tue May 03, 2011 8:01 am
by EViews Gareth
I'm not sure I understand your question.

Re: ARIMASel (Automatic ARIMA selection)

Posted: Tue May 03, 2011 8:22 am
by Sielenna
I have to determine the appropriate AR specification of at least 100 series. I want eviews to save the recommended specification under the same (or a slightly adjusted) name as the series. Is it possible to program this or do I as have to replace chosenstr (in your example) by the seriesname each time I run the program for a new series?

Re: ARIMASel (Automatic ARIMA selection)

Posted: Tue May 03, 2011 8:39 am
by EViews Gareth
Probably the best way to do it would be to change it in whatever program you're using that is calling the add-in.

Currently, with that modification, the add-in will create a string called "chosenstr" that contains the best specification. I assume you're calling the add-in in some sort of for-loop that runs through your 100 series. Thus after you have called the add-in, you should simply assign a new string (or whatever) to be equal to chosenstr. Thus it would look something like this:

Code: Select all

for %ser myseries1 myseries2 myseries3
{%ser}.arimasel
string {%ser}_chosen = chosenstr
next


Of course with 100, it might be better to store them in a table:

Code: Select all

!row = 1
table results
for %ser myseries1 myseries2 myseries3
{%ser}.arimasel
results(!row,1) = %ser
results(!row,2) = chosenstr
!row = !row + 1
next

Re: ARIMASel (Automatic ARIMA selection)

Posted: Wed May 04, 2011 1:49 am
by Sielenna
Thank you very much!

Re: ARIMASel (Automatic ARIMA selection)

Posted: Fri May 13, 2011 1:41 am
by Sielenna
Hi,

When I use the add-in to determine the optimal AR lag length of a series I get an AR(1) according to the Hannan-Quinn criteria.
When I estimate the same model as a VAR-object with the same series the Hannan-Quinn criteria points at an AR(5). I entered 10 lags as maximum lag length in both the Add-in and the automatic lag selection box. Why do I get different results?

Kind greetings,

Re: ARIMASel (Automatic ARIMA selection)

Posted: Fri May 13, 2011 7:53 am
by EViews Gareth
Why would you expect them to be the same?

Re: ARIMASel (Automatic ARIMA selection)

Posted: Mon May 16, 2011 12:18 am
by Sielenna
Because a VAR Object with only one endogenous variable and no exogenous variables is the same as an AR model of that variable with the same lags. (regressioncoefficients, standard deviations, t-values are all the same- so why wouldn't the optimal lag length be?)