Page 1 of 1

checking a ready model on out-of-sample data

Posted: Mon Apr 22, 2013 12:19 pm
by dzonijs
Hello!

My situation is following:
I am developing a GARCH model on stock returns. I have daily data of stock returns for a 4-year period.
1) I want to make a GARCH model based on observations of first 3-years. Let's say it's a Model X.
2) I want to apply the Model X to the 4th year data. (I want to see, how well this Model X fits to the 4th year data).

Actually the paragraph 1) is already completed for me - I have the Model X. my problem is paragraph 2) - how to fit Model X to 4th year data? and it would be great if I could also get statistical information like R^2, Adjusted R^2, Akaike, Schwarz etc. ...
Is it possible in EViews? and how to do it exactly?

Thanks in advance!

Re: checking a ready model on out-of-sample data

Posted: Mon Apr 22, 2013 12:59 pm
by EViews Gareth
You can forecast from your equation into the 4-year. That will get you fitted values, which you can then compare against the actual. EViews won't compute the out of sample R^2 or any other statistics, but, given the fitted values, you could compute the residuals and then calculate them by hand.

Re: checking a ready model on out-of-sample data

Posted: Tue Apr 23, 2013 6:38 am
by dzonijs
thank you.

but is it possible for GARCH model to somehow extract and show numeric values of conditional variance in EViews?
Right now I can only view the graph (View --> Garch Graph --> Conditional Variance)

Re: checking a ready model on out-of-sample data

Posted: Tue Apr 23, 2013 7:39 am
by EViews Gareth
When you forecast from a GARCH equation you have the option of forecasting the GARCH series too.

Re: checking a ready model on out-of-sample data

Posted: Fri Apr 26, 2013 7:51 am
by dzonijs
I have one more question.

I have original data with ~1000 observations. Then I estimate ARMA(1,1) model. When I click "View-->Actual, Fitted, Residual Table" then there are only ~700 values in this table. what could be the problem?

Re: checking a ready model on out-of-sample data

Posted: Fri Apr 26, 2013 7:54 am
by EViews Gareth
How many observations were used in the estimation (should say at the top of the estimation output).