EViews.com

I don't understand the question.

I am running the arimasel program in a loop and it works fine. Still, I have a few questions and would be happy if you could help.

1) Does the command allow to define the deterministic regressors? You can include a constant and trend if you want but since I am running the arimasel command in a loop with many variables, it would be nice that the deterministic regressors can be different for each arima model.

2) How can you change the command so that the windows with the best specification and the estimation output don't pop up? With many variables you have to click ok all the time and close all the windows afterwards. The do command in the arimasel code before equation doesn't seem to do the job.

3) For some of my time series I am missing the last or the two last observations because they are not yet published. So I want to estimate these series and forecast those observations. Do you know a way to program such that arimasel only applies on these series with the missing observations at the end?

Many thanks for your help!

Your best bet would be to write your own EViews program that does all of the things you want, using the ARIMASEL add-in as a starting place.

Hi I'm looking to loop arimasel over a list of series like some of the others before me;

I'm using the code

!row = 1
table results
for %ser c1 c2 c3
{%ser}.arimasel
results(!row,1) = %ser
results(!row,2) = chosenstr
!row = !row + 1
next


However I keep getting an error saying "arimasel is not a valid view for c1 in do_c1.arimasel". C1 is my series name for clarity.

Any ideas what's wrong?

Do you have the add-in installed properly?

Yes, it works fine when I use it for an individual series. Also in the manage add-ins section, it says installed.

Is there a proc assigned to the add-in properly?

I'm not sure on that. I might re-install the add-in to see if that works.

hi everybody.
i have two suggestion:
1-ARMAX: this Add-in select the best order of AR, MA, SAR, SMA, if i have exogenous variables Xi what is the best lag order for each variables to minimize AIC HQ and BIC. can you add an option to use max lag for exogenous variables up to 10 lags, it would be useful to chow the best model ARMAX. , and chow in ouput table the selected model.
2-critical value: how can chow the table by columns by 39 row for example.
I hope that you consider my suggestion for this add-in, unfortunately I'm not an expert in programming in EViews .

have a nice day for all.

I am using this plugin to select hundreds of series (I comment out the interactive diaglogs). And find that there is a handful of them, that the ARMA models do not achieve convergence.

It seems that, in the plugin, no checking whether the model achieves convergence.

Correct.

Better off using the built in routines instead. They'll also be considerably faster.

which is the built in routine?

Series.autoarma (ev9 only)

1-how can estimate best model without constant C, it can not do this.
2-when i use the Add-in to search best model it show table with the order lag selection like this (p,q)(P,Q), how can change the code to show it like this (p,d,q)(P,D,Q), why do you not put the regular and seasonal differencing? in X-13 options i can find it.

Because the procedure always tests an (0,0,0) model, it needs at least one regressor (such as a constant), otherwise there is nothing to estimate.
You could change the code so that it starts at a (1,0,1) model if you wanted by changing the AR/MA loops so that they start at 1 rather than 0.

Since the selection criteria have no impact on deciding the differencing level, the d,D terms in the table are redundant (they would be the same for every model), so they're dropped to keep the table neater.