EViews.com

hi.
I came across a problem and I just want to khow that how can I use "correlogram" in eviews to choose a best AR , MA or ARMA model and their best LAGs?

Just in the same way as you use ACF and PACF statistics/plots...

Hello,

I am trying to use arma model to make an out-of-sample forecast. To be exact I want to specify the best arma model!

I have in-sample 1781 observations and out-of-sample 514 (total 2295)

can you help me on how to do it in EViews?

thank you

I would like to ask if about the structure of the arma model

using least squares and 'sample' my dependent variable the structure for estimating different arma models

sample c ar(1) ma(1) ar(3) ma(3) : is this a correct structure?
sample c ar(1) ar(3) ma(1) ma(3) : is this the same as the above?
sample c ar(3) ma(3) ar(1) ma(1) : is this correct or have the ar to be in increasing order?

Thank you

The order doesn't matter. But it's unusual to have MA(3) or AR(3) without having MA(2) and AR(2).

The order doesn't matter. But it's unusual to have MA(3) or AR(3) without having MA(2) and AR(2).

So what do you mean? you have to always have in order like 1 2 3...

The order doesn't matter. But it's unusual to have MA(3) or AR(3) without having MA(2) and AR(2).

So what do you mean? you have to always have in order like 1 2 3...

You don't have to have them in order, but you probably should include all the intermediate orders.

Thanx a lot!!

Thanx a lot!!

Sorry to bother you again but if the structure does not matter how come that when I run
sample c AR(5) AR(4) AR(9) AR(1) MA(5) MA(4) MA(9) MA(1)

AND

sample c AR(1) AR(4) AR(5) AR(9) MA(1) MA(4) MA(5) MA(9)

gives me a different results and makes the coefficients not significant?

Sorry to bother you again but if the structure does not matter how come that when I run
sample c AR(5) AR(4) AR(9) AR(1) MA(5) MA(4) MA(9) MA(1)

AND

sample c AR(1) AR(4) AR(5) AR(9) MA(1) MA(4) MA(5) MA(9)

gives me a different results and makes the coefficients not significant?

Significant i mean the probabilities of the coefficients to be under 0,05. This must be that way in order to have a good forecast.

Hard to say without seeing the workfile.

Hello,

I am trying to use arma model to make an out-of-sample forecast. To be exact I want to specify the best arma model!

I have in-sample 1781 observations and out-of-sample 514 (total 2295)

can you help me on how to do it in EViews?

thank you

Sorry I had it uploaded before!

I use the 1781 observations for estimation and 2295 for the forecast

This is it

Sorry to bother you again but if the structure does not matter how come that when I run
sample c AR(5) AR(4) AR(9) AR(1) MA(5) MA(4) MA(9) MA(1)

AND

sample c AR(1) AR(4) AR(5) AR(9) MA(1) MA(4) MA(5) MA(9)

gives me a different results and makes the coefficients not significant?

Significant i mean the probabilities of the coefficients to be under 0,05. This must be that way in order to have a good forecast.

You might want to post a picture of your output. You should be getting the same results.

Ok now I am sending u the two tries.