In the same spirit, there is no reason to
exclude inside the loop (possibly many times)
when you can exclude outside the loop (just once).
I won't try to modify Lukas' original example code,
which is schematic.
But suppose you want to do something for each
distinct value of -group-, except that you
also need at least 13 valid observations on
some <varlist> to play some game.
egen any_missing = rowmiss(<varlist>)
bysort group any_missing : gen byte OK = (_N > 12) & (anymissing == 0)
egen newgroup = group(group) if OK
su newgroup, meanonly
forval i = 1/`r(max)' {
... if newgroup == `i'
...
}
Nick
[email protected]
Friedrich Huebler
> Lukas,
>
> See -help capture-. You can also add an -if- statement at the
> beginning of the loop that checks if the current model can be
> estimated.
>
> forvalues i = 1(1)5000 {
>
> * Move to next regression if condition is true
> if [condition, e.g. sample size too small] {
> continue
> }
>
> [other commands]
>
> }
Lukas Mohler <[email protected]> wrote:
> > I'm running several thousand regressions within a forvalues loop.
> > It goes like
> > this:
> >
> > forvalues 1(1)5000{
> >
> > ivreg2 ..... if selection==x_1, gmm robust
> >
> > x_1=x_1+1
> >
> > estout .....
> >
> > }
> >
> > For some values of selection I do not have enough observations to
> > estimate the
> > model. Also, other problems could arise with the data.
> >
> > This is why I'd like to suppress all possible error messages and
> > continue to loop through my regressions.
*
* 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/
