How do I assign more weight to certain variables?
Posted: Tue Sep 24, 2013 12:03 pm
by kjd31
.
Re: How do I assign more weight to certain variables?
Posted: Wed Sep 25, 2013 12:51 pm
by EViews Glenn
You can't do this with EViews least squares. And to be honest, I'm not entirely certain what you are asking for would mean in the context of least squares.
[edit] Just a note to indicate that the original questions has been deleted.
Re: How do I assign more weight to certain variables?
Posted: Fri Nov 15, 2013 7:24 am
by Dave_the_Forecaster
On a (not so) similar topic, I've been trying to devise a method by which I can assign more 'weight' to certain cross-sections in a panel regression. The idea being that when the coefficients are estimated, rather than using all cross-sections equally, the estimates are biased in such a way that the final model is more accurate for certain cross section at the expense of others.
The obvious one to try is to simply duplicate the relevant cross-sections in the data, so if I have 10 x-sections, and I want x-section Y to be twice as important as the others, I simply re-structure the data to have 11 x-sections, with the 11th being a copy of Y.
However, this solution causes issues when it comes to forecasting. I'd be more than grateful if anyone has any ideas about how I could achieve this, or can give me solid reasons why I shouldn't even be trying!
Cheers.