From smarkus@fas.harvard.edu Sun Mar 30 01:46:56 2003 From: smarkus@fas.harvard.edu To: gov2001@lists.gking.harvard.edu Subject: [gov2001-l] Feasible Generalized Least Squares Date: Sat, 29 Mar 2003 20:46:56 -0500 Message-ID: <001401c2f65e$3edaa370$9369f78c@stanman> MIME-Version: 1.0 Content-Type: multipart/mixed; boundary="===============8740298255747812014==" --===============8740298255747812014== Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit 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 Message
can = somebody,=20 please, explain what a "Feasible Generalized Least Squares" regression = is..? how=20 does one make it happen in R...?
 
thanks!
 

****************************

Stanislav=20 Markus
Ph.D. Candidate

Harvard University
Department of=20 Government

e: smarkus(a)fas.harvard.edu
t:= =20  617.513.5407

 
------=_NextPart_000_0015_01C2F634.56049B70-- --===============8740298255747812014==-- From kimai@fas.harvard.edu Sun Mar 30 14:01:29 2003 From: kimai@fas.harvard.edu To: gov2001@lists.gking.harvard.edu Subject: [gov2001-l] Feasible Generalized Least Squares Date: Sun, 30 Mar 2003 09:01:29 -0500 Message-ID: In-Reply-To: MIME-Version: 1.0 Content-Type: multipart/mixed; boundary="===============8024742615329150931==" --===============8024742615329150931== Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit 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 > --===============8024742615329150931==-- From smarkus@fas.harvard.edu Sun Mar 30 23:14:13 2003 From: smarkus@fas.harvard.edu To: gov2001@lists.gking.harvard.edu Subject: [gov2001-l] Feasible Generalized Least Squares Date: Sun, 30 Mar 2003 18:14:13 -0500 Message-ID: <004f01c2f712$13c19180$9369f78c@stanman> In-Reply-To: MIME-Version: 1.0 Content-Type: multipart/mixed; boundary="===============6706601490507124201==" --===============6706601490507124201== Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit 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(a)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 --===============6706601490507124201==-- From kimai@fas.harvard.edu Sun Mar 30 23:34:11 2003 From: kimai@fas.harvard.edu To: gov2001@lists.gking.harvard.edu Subject: [gov2001-l] Feasible Generalized Least Squares Date: Sun, 30 Mar 2003 18:34:11 -0500 Message-ID: In-Reply-To: <004f01c2f712$13c19180$9369f78c@stanman> MIME-Version: 1.0 Content-Type: multipart/mixed; boundary="===============9154001337060200944==" --===============9154001337060200944== Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit 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 --===============9154001337060200944==-- From king@harvard.edu Tue Feb 9 22:42:04 2021 From: Gary King To: gov2001@lists.gking.harvard.edu Subject: [gov2001-l] Feasible Generalized Least Squares Date: Sun, 30 Mar 2003 00:05:40 -0500 Message-ID: In-Reply-To: <001401c2f65e$3edaa370$9369f78c@stanman> MIME-Version: 1.0 Content-Type: multipart/mixed; boundary="===============1477151420831007127==" --===============1477151420831007127== Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit 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 > > > --===============1477151420831007127==--