5:40 a.m.
Dear All Gretl User,
I need some help with an regression analysis I am running right now. The
data consists of about 600 entities over 180 month and is organized as panel
data.
Here is my problem:
I want to run a fixed effect panel regression with two independent
(explanatory) variables. Gretl prints out the regression result and gives me
one estimated beta for each variable. However, in a next step I need a
estimated beta for each of the 600 entities in a list to do further
analysis.
My solution so far:
I could run 600 manual regressions and always write down the estimated beta,
but I hope there will be an easier way to handle this.
I would be very grateful for some help.
Kind regards,
Peter
9:16 a.m.
Dear Peter,
I don't know if you are familiar with gretls great scripting abilities.
I am not sure if I understand your problem, but I think you are looking for a
"loop".
Read some words about it in the command reference, or take a look at Allins recent example
with OLS (mail from 23/02/2010 on this list).
Find the short script below:
(Best way to learn is learning by doing!)
<script>
# create big list of all dependent series
list ylist = I*EX I*IM I*IT
# and the list of regressors
list xlist = # <- put your regressors here
# full results matrix, starts out empty
matrix Res = {}
loop foreach i ylist --quiet
# deflate the series
series $i_def = $i / def_us
# run the regression
ols $i_def 0 xlist --quiet
# grab the coefficients and standard errors
matrix res = $coeff' | $stderr'
# and add them to the big results matrix
Res = Res | res
endloop
# open a file, and write out the results
outfile "results.txt" --write
printf "%#12.6g", Res
outfile --close
</script>
Daniel
Peter Blömeke írta:
Dear All Gretl User,
I need some help with an regression analysis I am running right now.
The data consists of about 600 entities over 180 month and is
organized as panel data.
Here is my problem:
I want to run a fixed effect panel regression with two independent
(explanatory) variables. Gretl prints out the regression result and
gives me one estimated beta for each variable. However, in a next step
I need a estimated beta for each of the 600 entities in a list to do
further analysis.
My solution so far:
I could run 600 manual regressions and always write down the estimated
beta, but I hope there will be an easier way to handle this.
I would be very grateful for some help.
Kind regards,
Peter
------------------------------------------------------------------------
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
9:38 a.m.
Peter Blömeke schrieb:
Dear All Gretl User,
I need some help with an regression analysis I am running right now. The
data consists of about 600 entities over 180 month and is organized as
panel data.
Here is my problem:
I want to run a fixed effect panel regression with two independent
(explanatory) variables. Gretl prints out the regression result and
gives me one estimated beta for each variable. However, in a next step I
need a estimated beta for each of the 600 entities in a list to do
further analysis.
Hi Peter, I don't understand this last sentence. Assuming that for you
"beta" is the estimated coefficient vector, there is no separate beta
for each entity. The standard panel model has equal (homogeneous)
coefficients for all units. The only thing which is allowed to differ
between units are the constant terms. Do you want to recover the
estimated unit-specific constant terms?
cheers,
sven
12:16 p.m.
On Wed, 10 Mar 2010, Sven Schreiber wrote:
Peter Blömeke schrieb:
> I need some help with an regression analysis I am running right now. The
> data consists of about 600 entities over 180 month and is organized as
> panel data.
>
> Here is my problem:
>
> I want to run a fixed effect panel regression with two independent
> (explanatory) variables. Gretl prints out the regression result and
> gives me one estimated beta for each variable. However, in a next step I
> need a estimated beta for each of the 600 entities in a list to do
> further analysis.
Hi Peter, I don't understand this last sentence. Assuming that for you
"beta" is the estimated coefficient vector, there is no separate beta
for each entity. The standard panel model has equal (homogeneous)
coefficients for all units. The only thing which is allowed to differ
between units are the constant terms. Do you want to recover the
estimated unit-specific constant terms?
If it's the unit-specific intercepts you're after, then you want
the "accessor" $ahat. See the Function Reference under the Help
menu.
panel y 0 x1 x2
series ahat = $ahat
print ahat --byobs
Allin Cottrell
1:02 p.m.
So far I have I have developed my problem as follows:
My data consists of monthly stockreturns of 612 companies and corresponding
indexreturns of two stock indices. My data is organized as "stacked cross
section": All return of the first month of all 612 companies among each
other and then of the second month below that and so on. It is
110160x3-matrix.
Gretl should run 612 times one regression. So it would look roughly like
this: stockreturn = beta0 + beta1*indexreturn1 + beta2*indexreturn2 +
errorterm
In the end I need a 612x2-matrix with all beta1 and beta 2 coefficients.
Peter Blömeke
-----Ursprüngliche Nachricht-----
Von: gretl-users-bounces(a)lists.wfu.edu
[mailto:gretl-users-bounces@lists.wfu.edu] Im Auftrag von Allin Cottrell
Gesendet: Mittwoch, 10. März 2010 18:17
An: Gretl list
Betreff: Re: [Gretl-users] Panel Data, print out estimated betas
On Wed, 10 Mar 2010, Sven Schreiber wrote:
Peter Blömeke schrieb:
> I need some help with an regression analysis I am running right now. The
> data consists of about 600 entities over 180 month and is organized as
> panel data.
>
> Here is my problem:
>
> I want to run a fixed effect panel regression with two independent
> (explanatory) variables. Gretl prints out the regression result and
> gives me one estimated beta for each variable. However, in a next step I
> need a estimated beta for each of the 600 entities in a list to do
> further analysis.
Hi Peter, I don't understand this last sentence. Assuming that for you
"beta" is the estimated coefficient vector, there is no separate beta
for each entity. The standard panel model has equal (homogeneous)
coefficients for all units. The only thing which is allowed to differ
between units are the constant terms. Do you want to recover the
estimated unit-specific constant terms?
If it's the unit-specific intercepts you're after, then you want
the "accessor" $ahat. See the Function Reference under the Help
menu.
panel y 0 x1 x2
series ahat = $ahat
print ahat --byobs
Allin Cottrell
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
3:14 p.m.
On Wed, 10 Mar 2010, Peter Blömeke wrote:
So far I have I have developed my problem as follows:
My data consists of monthly stockreturns of 612 companies and corresponding
indexreturns of two stock indices. My data is organized as "stacked cross
section": All return of the first month of all 612 companies among each
other and then of the second month below that and so on. It is
110160x3-matrix.
Gretl should run 612 times one regression. So it would look roughly like
this: stockreturn = beta0 + beta1*indexreturn1 + beta2*indexreturn2 +
errorterm
In the end I need a 612x2-matrix with all beta1 and beta 2 coefficients.
If the index returns are the same in all regressions, you may want to
reshape your dataset like this
date company1 company2 ... company611 company612 index1 index2
... your data here ...
and then do
list Y = company*
list X = const index1 index2
matrix B = mols({Y}, {X})
matrix B = B[2:3,]'
of course, you'd have to reshape your data to a pure time-series format
(as opposed to panel), which is probabily going to be tedious, but having
done that, the rest should take you just a few seconds.
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
r.lucchetti(a)univpm.it
http://www.econ.univpm.it/lucchetti
4:12 p.m.
On Wed, 10 Mar 2010, Riccardo (Jack) Lucchetti wrote:
On Wed, 10 Mar 2010, Peter Bl�meke wrote:
> So far I have I have developed my problem as follows:
>
> My data consists of monthly stockreturns of 612 companies and corresponding
> indexreturns of two stock indices. My data is organized as "stacked cross
> section": All return of the first month of all 612 companies among each
> other and then of the second month below that and so on. It is
> 110160x3-matrix.
>
> Gretl should run 612 times one regression. So it would look roughly like
> this: stockreturn = beta0 + beta1*indexreturn1 + beta2*indexreturn2 +
> errorterm
>
> In the end I need a 612x2-matrix with all beta1 and beta 2 coefficients.
If the index returns are the same in all regressions, you may want to
reshape your dataset like this
date company1 company2 ... company611 company612 index1 index2
... your data here ...
and then do
list Y = company*
list X = const index1 index2
matrix B = mols({Y}, {X})
matrix B = B[2:3,]'
of course, you'd have to reshape your data to a pure time-series format
(as opposed to panel), which is probabily going to be tedious, but having
done that, the rest should take you just a few seconds.
Good idea, but not too tedious to reorganize a panel into distinct
time series. Here's an illustration, using one of the supplied
datasets.
<script>
# 6-firm panel, 15 time periods, 4 relevant variables
open greene14_1.gdt
# list the (4, in this case) variables we want
list X = 3 4 5 6
scalar nvars = 4
scalar nunits = 6
scalar T = 15
matrix M = mshape({X}, T, nvars*nunits)
nulldata T --preserve
k = 1
loop i=1..nvars -q
loop j=1..nunits -q
series x$i$j = M[,k]
k++
endloop
endloop
print --byobs
</script>
Now series x11 to x12 contain the separate time series for the
first original variable, series x21 to x26 contain the time series
for the second original variable, and so on.
Allin Cottrell
5362
days inactive
5362
days old
6 comments
6 participants
participants (6)
-
Allin Cottrell -
Kehl Dániel -
Peter Blömeke -
Peter Blömeke -
Riccardo (Jack) Lucchetti -
Sven Schreiber