29 Mar
2003
29 Mar
'03
8:46 p.m.
This is a multi-part message in MIME format.
------=_NextPart_000_0015_01C2F634.56049B70
Content-Type: text/plain;
charset="us-ascii"
Content-Transfer-Encoding: 7bit
can somebody, please, explain what a "Feasible Generalized Least
Squares" regression is..? how does one make it happen in R...?
thanks!
****************************
Stanislav Markus
Ph.D. Candidate
Harvard University
Department of Government
e: smarkus(a)fas.harvard.edu
t: 617.513.5407
------=_NextPart_000_0015_01C2F634.56049B70
Content-Type: text/html;
charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
<HTML><HEAD>
<META HTTP-EQUIV=3D"Content-Type" CONTENT=3D"text/html; =
charset=3Dus-ascii">
<TITLE>Message</TITLE>
<META content=3D"MSHTML 6.00.2723.2500" name=3DGENERATOR></HEAD>
<BODY>
<DIV><SPAN class=3D341474401-30032003><FONT face=3DArial size=3D2>can =
somebody,=20
please, explain what a "Feasible Generalized Least Squares" regression =
is..? how=20
does one make it happen in R...?</FONT></SPAN></DIV>
<DIV><FONT face=3DArial size=3D2></FONT> </DIV>
<DIV><SPAN class=3D341474401-30032003><FONT face=3DArial=20
size=3D2>thanks!</FONT></SPAN></DIV>
<DIV><SPAN class=3D341474401-30032003><FONT face=3DArial=20
size=3D2></FONT></SPAN> </DIV><!-- Converted from
text/plain format =
-->
<P align=3Dleft><FONT =
size=3D2>****************************<BR><BR>Stanislav=20
Markus<BR>Ph.D. Candidate<BR><BR>Harvard University<BR>Department
of=20
Government<BR><BR>e: <A=20
href=3D"mailto:smarkus@fas.harvard.edu">smarkus@fas.harvard.edu</A><BR>t:=
=20
617.513.5407</FONT> </P>
<DIV> </DIV></BODY></HTML>
------=_NextPart_000_0015_01C2F634.56049B70--
30 Mar
30 Mar
12:05 a.m.
a special case of this is weighted least squares. but there are other
special cases. it depends on what exactly you want to do.
Gary
On Sat, 29 Mar 2003, Stanislav Markus wrote:
can somebody, please, explain what a "Feasible
Generalized Least
Squares" regression is..? how does one make it happen in R...?
thanks!
****************************
Stanislav Markus
Ph.D. Candidate
Harvard University
Department of Government
e: smarkus(a)fas.harvard.edu
t: 617.513.5407
9:01 a.m.
If I remember correctly, FGLS is something like:
1. Run OLS
2. calculate residual variances
3. use these variances as weights and run weighted least squares.
Take a look at any econometrics textbook, you should be able to find it
out.
Kosuke
On Sun, 30 Mar 2003, Gary King wrote:
a special case of this is weighted least squares. but there are other
special cases. it depends on what exactly you want to do.
Gary
On Sat, 29 Mar 2003, Stanislav Markus wrote:
can somebody, please, explain what a
"Feasible Generalized Least
Squares" regression is..? how does one make it happen in R...?
thanks!
****************************
Stanislav Markus
Ph.D. Candidate
Harvard University
Department of Government
e: smarkus(a)fas.harvard.edu
t: 617.513.5407
_______________________________________________
gov2001-l mailing list
gov2001-l(a)fas.harvard.edu
http://www.fas.harvard.edu/mailman/listinfo/gov2001-l
6:14 p.m.
We are trying to replicate the results of a pooled regression which is
described as:
"Feasible Generalized Least Squares regressions run with Stata 6.0;
corrected for first-order autocorrelation using panel-specific process.
Each regression is run with dummy variables for every year (but one)
covered by the data."
We've been able to get very close results with Stata, but I'd like to
know how to do it in R as well..
Also, I'm not clear (a) what panel-specific process is, and (b) why
excluding one year "mitigates autocorrelation"? Does it matter which
year is excluded?
Thank you,
Stan
****************************
Stanislav Markus
Ph.D. Candidate
Harvard University
Department of Government
e: smarkus(a)fas.harvard.edu
t: 617.513.5407
-----Original Message-----
From: gov2001-l-admin(a)fas.harvard.edu
[mailto:gov2001-l-admin@fas.harvard.edu] On Behalf Of Gary King
Sent: Sunday, March 30, 2003 12:06 AM
To: Stanislav Markus
Cc: gov2001-l(a)fas.harvard.edu
Subject: Re: [gov2001-l] Feasible Generalized Least Squares
a special case of this is weighted least squares. but there are other
special cases. it depends on what exactly you want to do.
Gary
On Sat, 29 Mar 2003, Stanislav Markus wrote:
can somebody, please, explain what a "Feasible
Generalized Least
Squares" regression is..? how does one make it happen in R...?
thanks!
****************************
Stanislav Markus
Ph.D. Candidate
Harvard University
Department of Government
e: smarkus(a)fas.harvard.edu
t: 617.513.5407
_______________________________________________
gov2001-l mailing list
gov2001-l(a)fas.harvard.edu
http://www.fas.harvard.edu/mailman/listinfo/gov2001-l
6:34 p.m.
FGLS is a way to get around the heteroskedasticity in least squares. I
don't have any book with me right now (you need to take a look at any
econometrics textbook to get exact procedures). Basically FGLS is done
with the following 3 steps. In the first step, you run least squares and
get residuals. In the second step, you calculate, in this case, the first
order autocovariances of the residuals. In the final step, you use these
estimated autocovariances and run least squares. It should be
straightforward to do this in R.
Also, I'm not clear (a) what panel-specific
process is, and (b) why
excluding one year "mitigates autocorrelation"? Does it matter which
year is excluded?
(a) they are referring to the fact that there might be unobserved effects
specific to countries or individuals (whatever your units are).
(b) you can't include all the year dummy variables if you have an
intercept. so it doesn't matter which year is excluded.
Kosuke
7720
days inactive
7720
days old
gov2001@lists.gking.harvard.edu
4 comments
3 participants
participants (3)
-
Gary King -
kimai＠fas.harvard.edu -
smarkus＠fas.harvard.edu