 Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

# AW: st: Coefficients resulting from Cross-Sectional Regressions

 From Fabian Schönenberger To Subject AW: st: Coefficients resulting from Cross-Sectional Regressions Date Fri, 6 Apr 2012 19:43:21 +0200

```I try to explain the research idea. Based on the expression yid,t = a +
bid,t*xid,t + e I am looking for bid,t, where id = firm and t = firm years.
Y = lnebit, x = lnsales; for every firm on different amounts of yearly
observations. The coefficient b measures the sensitivity of ebit resulting
from business activity. I will further analyse the coefficients and try to
find out what financial policies (for instance equity-ratio) this dependence
creates. In order to do that I (think I) need b for each individual for
every point in time. Clustered b (on id) are not sufficient to reuse the
coefficients for regressions based on yearly observations. One way would be
to merge id with t, in order to get yearly coefficients. Doesn't this make
sense?

Fabian

-----Ursprüngliche Nachricht-----
Von: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] Im Auftrag von Nick Cox
Gesendet: Freitag, 6. April 2012 18:39
An: statalist@hsphsun2.harvard.edu
Betreff: Re: st: Coefficients resulting from Cross-Sectional Regressions

That sounds like a series of regressions based on precisely one data point
each.

And that makes no sense to me.

Nick

2012/4/6 Fabian Schönenberger <fabian.schoenenberger@gmail.com>:
> Then I am on the wrong track anyway. Which is the right formula to do
> panel regressions with ID and t, for every single ID on each point in
> time in order to get coefficients individualised by ID and t?
>
>
> -----Ursprüngliche Nachricht-----
> Von: owner-statalist@hsphsun2.harvard.edu
> [mailto:owner-statalist@hsphsun2.harvard.edu] Im Auftrag von Nick Cox
> Gesendet: Freitag, 6. April 2012 17:20
> An: statalist@hsphsun2.harvard.edu
> Betreff: Re: st: Coefficients resulting from Cross-Sectional
> Regressions
>
> statsby _b, by(gvkey) saving(coefsales): regress lnebit lnsales
>
> conducts separate regressions for each firm. The coefficient does not
> differ by year, as different years are pooled in each regression.
>
> You can put the results of -statsby- back in the original dataset by
> using -merge-.
>
> Alternatively,
>
> egen group = group(gvkey)
> gen coeff = .
> su group, meanonly
>
> qui forval i = 1/`r(max)'  {
>          regress lnebit lnsales if group == `i'
>          replace coeff = _b[lnsales] if group == `i'
> }
>
> Nick
>
> 2012/4/6 Fabian Schönenberger <fabian.schoenenberger@gmail.com>:
>
>> I am an absolute beginner in Stata and face now some challenges with
>> generating regression coefficients as new variable for further
> computations.
>> I use tsset id t to conduct cross-sectional regressions. ID is gvkey
>> which identifies firm (about 4?000 firms in my sample), t is fyear
>> which identifies the specific year (ranging from 1984 to 2010). What
>> I am looking for is the function to run cross-sectional regressions
>> and generating coefficients as new variable to my list of variables.
>> In other words: I am looking for b, specified for each ID for every t
>> (yi,t =
> ai + bi,t*xi,t).
>> I have tried the following:
>> statsby _b, by(gvkey) saving(coefsales): regress lnebit lnsales
>> However, two problems arise. First, the major problem, I only get
>> coefficients for gvkey, but not specified for every point in time
(fyear).
>> Second, the coefsales are safed in a new file, but I would prefer a
>> new variable in my list of variable.
>> Any suggestions to solve the two problems are highly appreciated.

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```