Notice: On March 31, it was **announced** that Statalist is moving from an email list to a **forum**. The old list will shut down on April 23, and its replacement, **statalist.org** is already up and running.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

From |
"Ma, Guang" <guang.ma@utdallas.edu> |

To |
"statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu> |

Subject |
RE: st: RE: RE: difference between robust and cluster option |

Date |
Sat, 17 Jul 2010 06:46:20 +0000 |

Do you mean you have 560*9=5040 or fewer observations in total? Or you have multiple observations for each firm-year? If the former, you cannot cluster by both firm and year. If the latter, you can define a new variable called "class" (for example) which concatenates firm and year ("IBM1990", for example). then use the following command: xtreg y x, fe vce(cluster class) nonest where "nonest" is a new option to specify that "panels are not nested within clusters". Hope it helps. Guang Ma ________________________________________ From: owner-statalist@hsphsun2.harvard.edu [owner-statalist@hsphsun2.harvard.edu] on behalf of Jing Zhou [jing.zhou@rmit.edu.au] Sent: Saturday, July 17, 2010 12:15 AM To: statalist@hsphsun2.harvard.edu Subject: AW: st: RE: RE: difference between robust and cluster option according to Petersen (2009) (estimating standard errors in finance panel data sets: comparing approaches, review of financial studies, 2009.), the standard errors clustered by firm and time can be a useful robustness check. My dataset is a sample of 560 firms over 9 years (unbalanced). Could you please advise me how I can cluster both firm and time in my analysis by using stata. cos when typing "xtreg, fe cluster (year)", the result shows "panels are not nested within clusters". Many thanks! Jing >>> "Martin Weiss" <martin.weiss1@gmx.de> 16/07/2010 9:18 pm >>> <> Maybe update 25feb2008 ... 20. xtreg, fe now uses vce(cluster id) when vce(robust) is specified, in light of the new results in Stock and Watson, "Heteroskedasticity-robust standard errors for fixed-effects panel-data regression," Econometrica 76 (2008): 155-174. HTH Martin -----Ursprüngliche Nachricht----- Von: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] Im Auftrag von Nils Braakmann Gesendet: Freitag, 16. Juli 2010 13:08 An: statalist@hsphsun2.harvard.edu Betreff: Re: st: RE: RE: difference between robust and cluster option As far as I know, using -robust- with a fixed effects estimator now automatically uses -cluster(id)- since some update in version 10.1 (might also be 10.0). The reason (again as far as I know) ist that Stock and Watson showed in an Econometrica-article in 2008 that the "normal" robust SEs are inconsistent with a FE-estimator (see James H. Stock and Mark Watson, 2008: “Heteroskedasticity-Robust Standard Errors for Fixed Effects Panel Data Regression”, Econometrica 76(1): 155-174). So, my guess would be that your results are identical because both of your approaches in fact use -cluster(id)- (and rightly so). Best, Nils On Fri, Jul 16, 2010 at 7:18 AM, Ma, Guang <guang.ma@utdallas.edu> wrote: > The coefficients estimated should be the same, since they are unbiased under both "robust" and "cluster", but the t-values and standard errors should differ. You may still get the same t-values, though, depending on your data and cluster variables. > > > Guang Ma > Accounting Department > The University of Texas at Dallas > School of Management, SM41 > Richardson, TX 75080 > > > > > > -----Original Message----- > From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Jing Zhou > Sent: Friday, July 16, 2010 12:09 AM > To: statalist@hsphsun2.harvard.edu > Subject: st: RE: difference between robust and cluster option > > thank you for your advisement. I tried cluster and robust in FE model individually, but the results are exactly the same. if "robust" only control for heteroskedasiticity while "cluster" for heteroskedasticity and serial correlation, why there is no difference between the two regressions? > >>>> "Ma, Guang" <guang.ma@utdallas.edu> 16/07/2010 2:54 pm >>> > Anyone is welcomed to correct me if I made any mistake. > > I think cluster can take good care of both heteroskedasticity and serial correlation, as long as you cluster in a panel way. "robust" is just heteroskedasticity consistent since it uses 1/(1-h)^2 as weights. Check this paper: Petersen, M. A. 2009. "Estimating standard errors in finance panel data sets: Comparing approaches." Review of Financial Studies 22 (1): 435-480. > > > Guang Ma > Accounting Department > The University of Texas at Dallas > School of Management, SM41 > Richardson, TX 75080 > > > > -----Original Message----- > From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Jing Zhou > Sent: Thursday, July 15, 2010 8:34 PM > To: statalist@hsphsun2.harvard.edu > Subject: st: difference between robust and cluster option > > To control for the heteroskedasticity and serial correlation in FE and RE models, do I need to add only "robust" or both "robust"and "cluster" option in the corresponding command? And after this remedy, is it unnecessary to worry about the two problems? > > > * > * 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/ > > > * > * 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/ > * * 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/ * * 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/

**Follow-Ups**:**RE: st: RE: RE: difference between robust and cluster option***From:*"Jing Zhou" <jing.zhou@rmit.edu.au>

**References**:**st: difference between robust and cluster option***From:*"Jing Zhou" <jing.zhou@rmit.edu.au>

**st: RE: difference between robust and cluster option***From:*"Ma, Guang" <guang.ma@utdallas.edu>

**st: RE: difference between robust and cluster option***From:*"Jing Zhou" <jing.zhou@rmit.edu.au>

**st: RE: RE: difference between robust and cluster option***From:*"Ma, Guang" <guang.ma@utdallas.edu>

**Re: st: RE: RE: difference between robust and cluster option***From:*Nils Braakmann <nilsbraakmann@googlemail.com>

**AW: st: RE: RE: difference between robust and cluster option***From:*"Martin Weiss" <martin.weiss1@gmx.de>

**AW: st: RE: RE: difference between robust and cluster option***From:*"Jing Zhou" <jing.zhou@rmit.edu.au>

- Prev by Date:
**st: Why does -centile- provide different results compared to -sum- or -pctile-?** - Next by Date:
**Re: st: Using mifit after ice to get imputed values** - Previous by thread:
**AW: st: RE: RE: difference between robust and cluster option** - Next by thread:
**RE: st: RE: RE: difference between robust and cluster option** - Index(es):