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

From |
John Antonakis <john.antonakis@unil.ch> |

To |
statalist@hsphsun2.harvard.edu |

Subject |
Re: st: RE: ivreg versus xtivreg |

Date |
Fri, 12 Feb 2010 23:16:20 +0100 |

Best, J. ____________________________________________________

Department of Organizational Behavior University of Lausanne Internef #618 CH-1015 Lausanne-Dorigny Switzerland Tel ++41 (0)21 692-3438 Fax ++41 (0)21 692-3305 Faculty page: http://www.hec.unil.ch/people/jantonakis Personal page: http://www.hec.unil.ch/jantonakis ____________________________________________________ On 12.02.2010 22:38, Schaffer, Mark E wrote:

John,-----Original Message-----From: owner-statalist@hsphsun2.harvard.edu[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf OfJohn AntonakisSent: 12 February 2010 21:02 To: statalist@hsphsun2.harvard.edu Subject: Re: st: RE: ivreg versus xtivregInterestingly, the Hansen J is non significant when I specifythe model with the "robust" vce option. However, I cannotobtain the J test when clustering (as mention, I get the"warming" message).I have a hunch ... there's an undocumented ivreg2 option -spsd-. For some special cases, it will take a non-pd VCV and make it pd using a spectral decomposition. What happens if you reestimate using this option? --MarkIntuitively, would you think that the Hansen J would be closeto that obtained from the two-way clustered J? I guess thatthere is no "trick"of sorts or something that could be done to get an idea ifthe overidentifying restrictions are viable.Best, J. ____________________________________________________ Prof. John Antonakis, Associate Dean Faculty of Business and Economics Department of Organizational Behavior University of Lausanne Internef #618 CH-1015 Lausanne-Dorigny Switzerland Tel ++41 (0)21 692-3438 Fax ++41 (0)21 692-3305 Faculty page: http://www.hec.unil.ch/people/jantonakis Personal page: http://www.hec.unil.ch/jantonakis ____________________________________________________ On 12.02.2010 17:53, Schaffer, Mark E wrote:Thanks, John. 2-way clustering is terra nova, and I aminterested in general about how it works in practice. Itsounds like your application is very suitable.Cheers, Mark-----Original Message----- From: owner-statalist@hsphsun2.harvard.edu[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of JohnAntonakisSent: Friday, February 12, 2010 4:04 PM To: statalist@hsphsun2.harvard.edu Subject: Re: st: RE: ivreg versus xtivreg Hi Mark:The SEs are quite similar (and still significant) using two-wayclustering as compared to the robust or clustering on onedimension.The first dimension has 192 clusters and the seconddimension has 99clusters (and observations are n =433). Regards, John. ____________________________________________________Prof. John Antonakis, Associate Dean Faculty of Business andEconomics Department of Organizational Behavior University ofLausanne Internef #618CH-1015 Lausanne-Dorigny Switzerland Tel ++41 (0)21 692-3438 Fax ++41 (0)21 692-3305 Faculty page: http://www.hec.unil.ch/people/jantonakis Personal page: http://www.hec.unil.ch/jantonakis ____________________________________________________ On 12.02.2010 14:26, Schaffer, Mark E wrote:John,-----Original Message----- From: owner-statalist@hsphsun2.harvard.edu[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of JohnAntonakisSent: Friday, February 12, 2010 12:03 PM To: statalist@hsphsun2.harvard.edu Subject: Re: st: RE: ivreg versus xtivreg Hi Mark: Yes; I am using the 2-way cluster option. Fyi, the highestn-size forthe cluster is 192.I take it that means you've got 192 clusters in onedimension. Do you have a similarly large number in the seconddimension? In that case I agree, you're probably OK.I'm also curious about the results of using 2-wayclustering. Does it make a big difference?--MarkGlad to hear that the SEs are probably OK. Best regards, John. ____________________________________________________Prof. John Antonakis, Associate Dean Faculty of Business andEconomics Department of Organizational Behavior University ofLausanne Internef #618CH-1015 Lausanne-Dorigny Switzerland Tel ++41 (0)21 692-3438 Fax ++41 (0)21 692-3305 Faculty page: http://www.hec.unil.ch/people/jantonakis Personal page: http://www.hec.unil.ch/jantonakis ____________________________________________________ On 12.02.2010 12:50, Schaffer, Mark E wrote:John,-----Original Message----- From: owner-statalist@hsphsun2.harvard.edu[mailto:owner-statalist@hsphsun2.harvard.edu] OnBehalf Of JohnAntonakis Sent: Friday, February 12, 2010 9:23 AM To: statalist@hsphsun2.harvard.edu Subject: Re: st: RE: ivreg versus xtivreg Hi Mark: I have a similar error; however, the message I get is:Warning: estimated covariance matrix of momentconditions not offull rank. overidentification statistic not reported, andstandard errors and model tests should be interpretedwith caution.Possible causes:number of clusters insufficient to calculate robustcovariance matrixsingleton dummy variable (dummy with one 1and N-1 0s orvice versa) partial option may address problem. Note, I have modeled two dimensions of clusteringDo you mean you're using the new 2-way cluster option?(and put all myfixed-effects in the partial option). Thus, I am correct insaying that the z test for a particular coefficientestimate isstill interpretable.Probably. I can think of some special cases where thingscan go wrong (these are what motivated the writing ofthe warningmessage). You might have a small number of clusters (so theasymptotics are dodgy) or a singleton dummy which would mean youcouldn't do a test on the significance of the coefficient (ofcourse, you couldn't do that anyway). But most likelythe SEs areOK.--MarkBest, J. ____________________________________________________Prof. John Antonakis, Associate Dean Faculty of Business andEconomics Department of Organizational Behavior University ofLausanne Internef #618CH-1015 Lausanne-Dorigny Switzerland Tel ++41 (0)21 692-3438 Fax ++41 (0)21 692-3305 Faculty page: http://www.hec.unil.ch/people/jantonakis Personal page:http://www.hec.unil.ch/jantonakis____________________________________________________On 12.02.2010 09:55, Schaffer, Mark E wrote: > Sara, > >> -----Original Message----- >> From: owner-statalist@hsphsun2.harvard.edu>> [mailto:owner-statalist@hsphsun2.harvard.edu] OnBehalf Ofsara borelli >> Sent: 12 February 2010 07:58 >> To:statalist@hsphsun2.harvard.edu >> Cc: Schaffer, Mark E >>Subject: R: st: RE: ivreg versus xtivreg >> >> HiMark, >> Iam sorry I did not understand your answer...and Ithink it >> isbecause I I made a typo in writing the command and Idid >> notexplain myself clearly. I do NOT include i.county in >>xtivreg2. So I reformulate my question:>>>> xtivreg2 CRit (Xit=Wit) Zit (state by yearfixed effects),>> fe i(county), cluster(county) >> >> gives theerror message" estimated covariance matrix of >> moment conditions not offull rank; overidentification >> statistic not reported, andstandard errors and model tests >> should beinterpreted withcaution. Possible causes:>> singleton dummy variable (dummy with one 1 and N-10s or viceversa) fwl option may address problem">>>> but the same command xtivreg2 without state byyear effects:>> xtivreg2 CRit (Xit=Wit) Zit, fe i(county),cluster(county)does NOT report the above error message >> >> I do notuderstand why the inclusion of state by year fe >>caused thaterror message > > I think your explanation in your earlieremail isprobably right:> >> I noticed that 2 states have a number of counties(clusters) lower>> than the number of years available. By dropping these twostates STATA >> run the xtivreg2 without giving the errormessage.>>>> I have read the help file and FAQ and I think something isgoing with >> those state by year effects thatcreates a kind ofsingleton dummy>> problem, but I am not sure Any help would beappreciated >And the question is, should you worry about it? TheVCV shouldstill be OK for tests of, say, one parameter at atime. So longas you arearen't trying to do something that requires afull-rank VCV (like2-step GMM, or using an overid stat, or joint testing of allthe coeffs),you're OK, I think. > > Memo to self: calling this an "error" in the outputis possiblyoverstating the issue; maybe "warning" is better. > > --Mark > >> thank you >> sara >> >> >>>> --- Gio 11/2/10, Schaffer, Mark E<M.E.Schaffer@hw.ac.uk> hascritto: >>>>> Da: Schaffer, Mark E <M.E.Schaffer@hw.ac.uk> >>>Oggetto:st: RE: ivreg versus xtivreg >>> A:statalist@hsphsun2.harvard.edu >>> Data: Giovedì 11 febbraio2010, 19:43 >>> Sara, >>> >>>> -----Original Message-----From: owner-statalist@hsphsun2.harvard.edu>>>> [mailto:owner-statalist@hsphsun2.harvard.edu] >>> On Behalf Of >>>> sara borelli>>>> Sent: Thursday, February 11, 2010 6:32 PM >>>> To:statalist@hsphsun2.harvard.edu >>>> Subject: st: ivreg versusxtivreg >>>> >>>> Dear members, >>>> I am running thefollowing regressions in STATA8.2:>>>> CRit = a*Xit + b*Zit + (state by >>> year fixedeffects) + >>>> (county fixed effects),cluster(county) >>>>CRit= crime rate in county i year t >>>> Xit =endogenousregressor >>>> Zit = set of 7 exogenous regressorsstateby year fixed effects = interactions between >>> indicatorsfor each state and year >>>> >>>> I instrumentXit usingWit >>>> There are 246 counties, 10 states, 8 years Ihave beenrunning this >>>> model with two commands >>>>>>>> xi: ivreg CRit >>> (Xit=Wit) Zit (state by year fixed>>>> effects) i.county, cluster(county) >>>>xtivreg2 CRit(Xit=Wit) >>> Zit (state by year fixed>>>> effects) i.county, fe i.(county) >>> cluster(county)the two commands give exactly the samecoefficient, >>> andjust >>>> slightly different std errors, but afterxtivreg2 Iget the >>>> following message:>>>> Error: estimated covariance matrix of moment >>>conditions not >>>> of full rank; >>>> overidentificationstatistic not reported, and >>> standard>>>> errors and model tests >>> should be interpreted with>>>> caution. Possible causes: singleton dummy >>> variable(dummy >>>> with one 1 and N-1 0s or vice versa) fwloption mayaddress problem.>>>> I noticed that 2 states have a number of counties >>>(clusters) >>>> lower than the number of years available. Bydropping >>> these >>>> two states STATA run the xtivreg2without giving the >>> error message.>>>> I have read the help file and FAQ and I think >>>something is >>>> going with those state by year effects thatcreates a >>> kind of >>>> singleton dummy problem, but I amnot sure Anyhelp would be>>>> appreciated >>> It's not a problem. >>>>>> You will note that xtivreg2 does not report thefixed effectdummy >>> variables. >>>>>> xi: ivreg2 explicitly includes the dummies, and the >>coefficients are >>> reported.>>>>>> The warning message reported by ivreg2 istriggered by thefact that >>> there are now many more rows/columns in theVCV, the new, >> extended VCV >>> is no longer full rank. >>>>>> In fact, the coefficients and standard errors for the >>dummies aren't >>> consistent anyway (under the usualpanel dataassumptions - this is >>> the "incidental parametersproblem").So you really aren't >> interested >>> in the extended VCVanyway.>>> >>> HTH, >>> Mark >>> >>>> thank you >>>> sara >>>> >>>> >>>>>>>>>>>>>>>> * >>>> * 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/ >>>> >>> >>> -->>> Heriot-Watt University is a Scottish charityregistered >>under charity >>> number SC000278. >>> >>> >>> * >>> * 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/

**References**:**st: RE: ivreg versus xtivreg***From:*"Schaffer, Mark E" <M.E.Schaffer@hw.ac.uk>

**R: st: RE: ivreg versus xtivreg***From:*sara borelli <saraborelli77@yahoo.it>

**RE: st: RE: ivreg versus xtivreg***From:*"Schaffer, Mark E" <M.E.Schaffer@hw.ac.uk>

**Re: st: RE: ivreg versus xtivreg***From:*John Antonakis <john.antonakis@unil.ch>

**RE: st: RE: ivreg versus xtivreg***From:*"Schaffer, Mark E" <M.E.Schaffer@hw.ac.uk>

**Re: st: RE: ivreg versus xtivreg***From:*John Antonakis <john.antonakis@unil.ch>

**RE: st: RE: ivreg versus xtivreg***From:*"Schaffer, Mark E" <M.E.Schaffer@hw.ac.uk>

**Re: st: RE: ivreg versus xtivreg***From:*John Antonakis <john.antonakis@unil.ch>

**RE: st: RE: ivreg versus xtivreg***From:*"Schaffer, Mark E" <M.E.Schaffer@hw.ac.uk>

**Re: st: RE: ivreg versus xtivreg***From:*John Antonakis <john.antonakis@unil.ch>

**RE: st: RE: ivreg versus xtivreg***From:*"Schaffer, Mark E" <M.E.Schaffer@hw.ac.uk>

- Prev by Date:
**RE: st: RE: ivreg versus xtivreg** - Next by Date:
**st: xtmixed: How replicate a SAS example in Stata11** - Previous by thread:
**RE: st: RE: ivreg versus xtivreg** - Next by thread:
**st: nnmatch: Use only some observables for matching, and the others as regression controls?** - Index(es):

© Copyright 1996–2016 StataCorp LP | Terms of use | Privacy | Contact us | What's new | Site index |