Re: st: my code for event study

["Michael Blasnik" <michael.blasnik@verizon.net>]

I haven't looked that closely at how the dataset is structured, but I think I'd 
try Ingo's suggestion first -- the statsby code uses the -in- approach for 
sample identification and so works pretty quickly.  Something like this may do 
what you want:

keep if est_window==1
statsby , by(id): reg ret mkt
sort id
save mycoeffs
use mydata
sort id
merge id using mycoeffs
gen pred_rtn=_b_cons+_b_mkt*mkt if event_window==1

Some other suggestions that will speed up the code you show include making sure 
to drop all observations that aren't needed -drop if est_window!=1 & 
event_window!=1, and even breaking up the dataset into smaller pieces and run 
the loop on each chunks and re-assemble the results.

Michael Blasnik

----- Original Message ----- 
From: "Ingo Brooks" <ingo.brooks@gmail.com>
To: <statalist@hsphsun2.harvard.edu>
Sent: Thursday, March 15, 2007 4:03 AM
Subject: Re: st: my code for event study



> Nian,
>
> Did you also try to use -statsby- , then merge the coefficient
> estimates to your existing dataset and compute the predicted values by
> yourself? I think this should be much faster and easier to implement
> than the "in" solution.
>
> Best,
> Ingo
>
> On 3/15/07, Nian Huang <huangnian@gmail.com> wrote:
> > I want to thank everybody for their previous quick responses.
> >
> > Here is my piece of code for event study. Basically, I get it from the
> > Stata sample program.
> >
> >
> > forvalues i=1(1)1314 {
> >         quietly reg ret mkt if id==`i' & est_window==1
> >         quietly predict p if id==`i'
> >         quietly replace pred_rtn= p if id==`i' & event_window==1
> >         quietly drop p
> > }
> >
> > The above program took 30 minutes to run through. Is there any way to
> > improve it?
*
*   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/