I haven't tried to understand your Stata
code, because from your word specification
the problem seems to yield to a direct climb
up the North Face.
There will be as many regressions as observations.
I am not clear if your -Id- is equal
to your observation numbers, but we can go
forval i = 1/`=_N' {
regress <varlist> if _n != `i' & Country == Country[`i']
<do something with results>
}
However, your various regressions are highly dependent
within countries, which will be problematic for
some purposes. Whether your approach is correct
I cannot judge without knowing how you intend
to use the results.
Nick
[email protected]
Apostolos Ballas
> I am interested to do a statistical analysis repeteadly for a
> number of
> companies. The observations are organized as follows:
>
> Country Id Numerical Data
> UK 1
> UK 2
> UK 3
> .
> .
> .
> France 101
> France 102
> .
>
> and so on for other EU countries.
>
> I want to run a regression from which I will exclude company
> X but use data
> from all other companies in the same country. My solution is
> as follows:
>
> gen byte iscountry=0
> sort id year
> gen obs=_n
> by id: gen obs_no=_n
> by id: gen obs_lot=_N
> sort obs_no id
> gen comp_no=_n if obs_no==1
> sort obs
> replace comp_no=comp_no[_n-1] if comp_no >=.
> forval i=1/1302 {
> replace iscountry=0
> replace iscountry=1 if comp_no==`i'
> bysort country (iscountry): replace iscountry = iscountry[_N]
> sureg ....... if comp_no!=`i' & iscountry==1, robust
> }
>
> Is this the best possible approach? Indeed, is it a correct approach?
>
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
