panel data: removing data cross-sectional individuals

[helplessbunny]

!detect and eliminate constant cross-sectional individuals!

Hi!

I am working with a weakly balanced panel and I want to do a ABI and II estimation which uses FD for IV generation.

My dependent variable is binary. Therefore I want to eliminate all cross-sectional individual that are constant in all time periods. (my Prof told me to do that before I proceed with FD)

So if f.e. individual 1 is 0 in all periods, I want to remove the individual not only form the dependent variable, but from the whole data set, ergo all 7 colums. Individual 2 changes from 0 to 1 half way through the periods, so all periods should be spared.

Is this possible? Did I overlook something in the manual?

Sincerely

[EViews Gareth]

Set the sample to be:

Code: Select all

if @varsby(Y, @crossid)<>0

where Y is the name of your dependent variable.

@varsby(Y, @crossid) calculates the variance of Y by cross-section. If Y takes the same value for each period for a cross-section, its variance will be 0. So simply set the sample to be any observations where the variance isn't 0.