Page 1 of 1

Sampling in a 3D unbalanced Panel

Posted: Tue Apr 23, 2013 1:34 pm
by gussy
Hi there!

I'm having trouble setting the right sampling command for my dataset, which is specified by a year date identifier and two cross-sectional identifiers (ISIC and country). Now, I want to sample my data but only for certain years for certain countries
e.g.
Brazil 1995
Peru 1995-1997
Chile 1995-2002
etc.

I've currently only managed to do this by opening up my Excel data as an unstructured panel then using a very long IF statement; I've found that I can't include year commands in the IF statement box.

Any advice?

P.S. I've come across a couple of posts which mention that 'Eviews does not support panel estimation methods for 3D panels'. Does this mean that estimation will still be done correcly by Eviews with 3D panels, although I can't specify fixed/ random effects for example, or that estimation doesn't really work at all for the 3D panel?

Thanks!

Re: Sampling in a 3D unbalanced Panel

Posted: Tue Apr 23, 2013 5:41 pm
by EViews Gareth
Estimation will work, just no fixed/random effects.

You'll have to use a long if condition. Something like:

Code: Select all

smpl if (country="Brazil" and @year=1995) or (country="Peru" and @year>=1995 and @year<=1997) or ....

Re: Sampling in a 3D unbalanced Panel

Posted: Sun Apr 28, 2013 2:51 am
by gussy
Brilliant!

Works perfectly, thank you for the quick reply!

I have another question also:

I'm having trouble computing robust covariances in my model where I move to including a specific (binary) independent variable (which is pretty collinear with the (binary) dependent variable) and a set of year dummy variables from not including either. I get an error message of "Triangular matrix too small or source matrix asymmetric".

Does this mean that the dependent variable doesn't vary 'enough' within the new sample => robust covariances can't be computed?

Thanks again