optimal Lags in ARLI Model

[CDN123]

Hi
I am a new eviews user and still becoming familiar with the program.

I would like create an autoregressive leading indicator model, which contains lagged terms of the dependent variable (the variable being forecasted) and lagged terms of the leading indicator.

Is there a program that can be used to quickly determine the optimal number of lags for both the dependent variable and the leading indicator using the Akaike information criteria.

Thanks

Mike

[trubador]

http://www.eviews.com/help/helpintro.ht ... .2.html%23

[CDN123]

Thanks. I found the option for the ARDL model. I would only like to include lagged values of the indicator and it seems to want to include a t=0 term as well.

Also, is it possible to estimate this model with all the variables in first differences and then forecast in levels. My data is not stationary.

Thanks

Mike

[trubador]

Thanks. I found the option for the ARDL model. I would only like to include lagged values of the indicator and it seems to want to include a t=0 term as well.

From the practical point of view, you should also test the contemporaneous relationship to see if the regressor is a true leading indicator. Still, if you do not want to include lag zero, just use the first lag of your regressor as the explanatory variable (e.g. y c x(-1)) and then set the Max lags option for automatic selection with AIC.

Also, is it possible to estimate this model with all the variables in first differences and then forecast in levels. My data is not stationary.

Yes, it is possible. However, ARDL model does not require stationarity. You can use Bounds test to decide between I(0) and I(1) and see if any long-run relationships are present.

[CDN123]

Thanks this is great. I understand now.

One last thing. I want to have my data in first differences for the forecast. Can I enter my dependent and independent variables in first differences into the automatic ARDL option and determine the max number of lags and then use the forecast button to see the forecast in levels.

How would I do this?

Sorry I am new to the program.

[trubador]

You can use the difference operator in the equation window (e.g. d(y), d(x(-1))). You should at least have a look at Help->PDF Docs->EViews Illustrated before going any further.

[CDN123]

Great. Thank you very much for all your help.