Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: st: RE: non full ranked covariance matrix of moment conditions after xtivreg2
From
"Schaffer, Mark E" <[email protected]>
To
"[email protected]" <[email protected]>
Subject
RE: st: RE: non full ranked covariance matrix of moment conditions after xtivreg2
Date
Thu, 2 Jan 2014 01:55:28 +0000
Anat,
No problem. FYI I found the bug. It affected only the reporting; the VCEs are fine.
BTW, I see you're using the Bartlett kernel with bandwidth=1. This won't do what you want, because it means that zero lags are actually used (in the Bartlett kernel, the weights go linearly to zero and technically the last lag in the bandwidth gets a zero weight). With 17 years of data, the automatic bandwidth selection in ivreg2 and ivregress suggests a bandwith of 14 with the Bartlett kernel. You have a panel, and the auto selection is for pure time series, but at least it's something to go on.
--Mark
> -----Original Message-----
> From: [email protected] [mailto:owner-
> [email protected]] On Behalf Of Anat (Manes) Tchetchik
> Sent: 02 January 2014 01:22
> To: [email protected]
> Subject: Re: st: RE: non full ranked covariance matrix of moment conditions
> after xtivreg2
>
> Thank You!
>
> On Thu, Jan 2, 2014 at 2:34 AM, Schaffer, Mark E <[email protected]>
> wrote:
> > Anat,
> >
> > I'm glad the partialling out solved the problem.
> >
> > I can also reproduce the strange reporting bug about the number of clusters
> in the 2nd group. I'll look into fixing that bug shortly. Thanks for reporting it!
> >
> > Best wishes,
> > Mark
> >
> >> -----Original Message-----
> >> From: [email protected] [mailto:owner-
> >> [email protected]] On Behalf Of Anat (Manes) Tchetchik
> >> Sent: 02 January 2014 00:07
> >> To: [email protected]
> >> Subject: Re: st: RE: non full ranked covariance matrix of moment
> >> conditions after xtivreg2
> >>
> >> Dear Mark,
> >> I placed each answer below the relevant question for ease of following:
> >>
> >> > Can you check that you have the latest versions of ivreg2, xtivreg2
> >> > and
> >> ranktest installed, and also tell us which version of Stata you are using?
> >>
> >> > which ivreg2, all
> >> c:\ado\plus\i\ivreg2.ado
> >> *! ivreg2 3.1.07 28Jul2013
> >> *! authors cfb & mes
> >>
> >> > which xtivreg2, all
> >> c:\ado\plus\x\xtivreg2.ado
> >> *! xtivreg2 1.0.13 28Aug2011
> >> *! author mes
> >>
> >> > which ranktest, all
> >> c:\ado\plus\r\ranktest.ado
> >> *! ranktest 1.3.02 27Feb2012
> >> *! author mes, based on code by fk
> >>
> >> > version
> >> version 13.1
> >>
> >> > Does xtivreg2 report that any observations are not used because of
> >> singletons?
> >> No- it stated among the possible causes: singleton dummy variable
> >> (dummy with one 1 and N-1 0s or vice versa)
> >>
> >> > What happens if you estimate using ivreg2 and explicit dummies for
> >> countries?
> >>
> >> If I understand you correct, I ran:
> >> xi: ivreg2 SS dem gini inflation shad_econ_min3 gdp i.year
> >> dcountry_code1 ......... dcountry_code34 , cluster( country_code
> >> year) bw(1)
> >> the results remain the same , the same coefficients,warnings and the:
> >> Number of clusters (year) = 2007 (I can send you the results)
> >>
> >> > What happens if you estimate using xtivreg2 and partial-out the
> >> > year
> >> dummies?
> >>
> >> This has actually solved the problem as you can see below (yet it
> >> still says Number of clusters (year) = 2007):
> >> I ran the following:
> >> xi: xtivreg2 ss demgini gini inflation shad_econ sixtfive gdp
> >> i.year, fe cluster( country_code year) bw(1) partial(i.year) and received:
> >> i.year _Iyear_1989-2007 (naturally coded; _Iyear_1989 omitted)
> >>
> >> FIXED EFFECTS ESTIMATION
> >> ------------------------
> >> Number of groups = 35 Obs per group: min = 8
> >> avg = 12.6
> >> max = 16
> >>
> >> OLS estimation
> >> --------------
> >>
> >> Estimates efficient for homoskedasticity only Statistics robust to
> >> heteroskedasticity and clustering on country_code and year and
> >> kernel-robust to common correlated disturbances (Driscoll-Kraay)
> >> kernel=Bartlett; bandwidth=1
> >> time variable (t): year
> >> group variable (i): country_code
> >>
> >> Number of clusters (country_code) = 35 Number of obs = 440
> >> Number of clusters (year) = 2007 F( 6, 34) = 4.40
> >> Prob > F = 0.0022
> >> Total (centered) SS = 2.545869056 Centered R2 = 0.1594
> >> Total (uncentered) SS = 2.545869056 Uncentered R2 = 0.1594
> >> Residual SS = 2.139947433 Root MSE = .07269
> >>
> >> --------------------------------------------------------------------------------
> >> | Robust
> >> ss | Coef. Std. Err. z P>|z| [95%
> >> Conf. Interval]
> >> ---------------+-----------------------------------------------------
> >> ---------------+---
> >> ---------------+--------
> >> dem_gini | .0049295 .001544 3.19 0.001 .0019034 .0079556
> >> gini_m | -.005013 .0024799 -2.02 0.043 -.0098736 -.0001524
> >> inflation | -.0000572 .0000218 -2.62 0.009 -.0001 -.0000145
> >> shad_eco | -.0132554 .0053642 -2.47 0.013 -.0237691 -.0027416
> >> sixtfive | -.0003735 .0149491 -0.02 0.980 -.0296733 .0289262
> >> gdp | -.0085065 .0098609 -0.86 0.388 -.0278335 .0108204
> >> ---------------------------------------------------------------------
> >> ----------- Included instruments: demgini_min2 gini inflation
> >> shad_eco
> >> sixtfive gdp
> >> Partialled-out: _Iyear_1991 _Iyear_1992 _Iyear_1993 _Iyear_1994
> >> _Iyear_1995 _Iyear_1996 _Iyear_1997 _Iyear_1998
> >> _Iyear_1999 _Iyear_2000 _Iyear_2001 _Iyear_2002
> >> _Iyear_2003 _Iyear_2004 _Iyear_2005 _Iyear_2006
> >> _Iyear_2007
> >> nb: small-sample adjustments account for
> >> partialled-out variables
> >> ---------------------------------------------------------------------
> >> ---------
> >>
> >> > --Anat
> >> >
> >> >> -----Original Message-----
> >> >> From: [email protected] [mailto:owner-
> >> >> [email protected]] On Behalf Of Anat (Manes)
> >> >> Tchetchik
> >> >> Sent: 01 January 2014 20:09
> >> >> To: [email protected]
> >> >> Subject: st: non full ranked covariance matrix of moment
> >> >> conditions after
> >> >> xtivreg2
> >> >>
> >> >> Dear all,
> >> >>
> >> >> I have an unbalanced panel of data on 35 countries over 17 years
> >> >> (446
> >> >> obs.) I have jointly significant time effects are, as well as
> >> >> first order
> >> autocorrelation.
> >> >> As I want to account for two-way (unit and time ) cluster-robust
> >> >> covariance matrix and estimated the following model:
> >> >>
> >> >> xi: xtivreg2 SS dem gini inflation shad_econ_min3 gdp i.year,
> >> >> fe cluster( country year) bw(1)
> >> >>
> >> >> The output below raises two issues:
> >> >> (1) Why does it indicate that the "Number of clusters (year) = 2007"
> >> >> and not 17 as it should?
> >> >> (2) how should I treat the Warning: "estimated covariance matrix
> >> >> of moment conditions not of full rank. model tests should be
> >> >> interpreted with
> >> caution"
> >> >> Note that when omitting the year dummies this warning doesn't
> >> >> appear
> >> >>
> >> >> Number of clusters (country) = 35 Number of obs = 440
> >> >> Number of clusters (year) = 2007
> >> >>
> >> >> ss Coef. Std. Err.
> >> >> ---------------+-------------------------------------------------
> >> >> dem 0.00493 0.001544 3.19
> >> >> gini -0.005013 0.00248 -2.02
> >> >> inflation -5.72E-05 2.18E-05 -2.62
> >> >>
> >> >>
> >> >>
> >> >> gdp -0.008507 0.009861 -0.86
> >> >> _Iyear_1991 0.138 0.038671 3.56
> >> >> _Iyear_1992 0.365 0.079991 4.56
> >> >> _Iyear_1993 0.332 0.081221 4.09
> >> >> _Iyear_1994 0.339 0.081055 4.18
> >> >> _Iyear_1995 0.338 0.084453 4.01
> >> >> _Iyear_1996 0.360 0.088365 4.07
> >> >> _Iyear_1997 0.371 0.093862 3.96
> >> >> _Iyear_1998 0.363 0.100767 3.6
> >> >> _Iyear_1999 0.400 0.106786 3.75
> >> >> _Iyear_2000 0.422 0.10945 3.86
> >> >> _Iyear_2001 0.418 0.115567 3.62
> >> >> _Iyear_2002 0.416 0.12068 3.45
> >> >> _Iyear_2003 0.415 0.124553 3.33
> >> >> _Iyear_2004 0.418 0.128937 3.24
> >> >> _Iyear_2005 0.418 0.129226 3.23
> >> >> _Iyear_2006 0.475 0.133707 3.55
> >> >> _Iyear_2007 0.519 0.14433 3.59
> >> >>
> >> >>
> >> >> Warning: estimated covariance matrix of moment conditions not of
> >> >> full
> >> rank.
> >> >> model tests should be interpreted with caution.
> >> >> Possible causes:
> >> >> number of clusters insufficient to calculate robust covariance
> matrix
> >> >> covariance matrix of moment conditions not positive definite
> >> >> covariance matrix uses too many lags
> >> >> singleton dummy variable (dummy with one 1 and N-1 0s or
> >> >> vice versa) partial option may address problem.
> >> >>
> >> >> Best,
> >> >>
> >> >> Anat Tchetchik, PhD
> >> >> Department of Business Administration Guilford Glazer Faculty of
> >> >> Business and Management Ben-Gurion University of the Negev
> >> >> P.O.Box: 653
> >> >> Beer-Sheva, Israel, 84105
> >> >>
> >> >> E-mail: [email protected]
> >> >> Phone 972-(0)8-6479735
> >> >> Fax: 972-(0)8-6472920
> >> >> *
> >> >> * For searches and help try:
> >> >> * http://www.stata.com/help.cgi?search
> >> >> * http://www.stata.com/support/faqs/resources/statalist-faq/
> >> >> * http://www.ats.ucla.edu/stat/stata/
> >> >
> >> >
> >> > -----
> >> > Sunday Times Scottish University of the Year 2011-2013 Top in the
> >> > UK for student experience Fourth university in the UK and top in
> >> > Scotland (National Student Survey 2012)
> >> >
> >> > We invite research leaders and ambitious early career researchers
> >> > to join us in leading and driving research in key inter-disciplinary themes.
> >> > Please see www.hw.ac.uk/researchleaders for further information and
> >> > how to apply.
> >> >
> >> > Heriot-Watt University is a Scottish charity registered under
> >> > charity number SC000278.
> >> >
> >> >
> >> > *
> >> > * For searches and help try:
> >> > * http://www.stata.com/help.cgi?search
> >> > * http://www.stata.com/support/faqs/resources/statalist-faq/
> >> > * http://www.ats.ucla.edu/stat/stata/
> >>
> >>
> >>
> >> --
> >> Anat Tchetchik, PhD
> >> Department of Business Administration Guilford Glazer Faculty of
> >> Business and Management Ben-Gurion University of the Negev
> >> P.O.Box: 653
> >> Beer-Sheva, Israel, 84105
> >>
> >> E-mail: [email protected]
> >> Phone 972-(0)8-6479735
> >> Fax: 972-(0)8-6472920
> >> *
> >> * For searches and help try:
> >> * http://www.stata.com/help.cgi?search
> >> * http://www.stata.com/support/faqs/resources/statalist-faq/
> >> * http://www.ats.ucla.edu/stat/stata/
> >
> >
> > -----
> > Sunday Times Scottish University of the Year 2011-2013 Top in the UK
> > for student experience Fourth university in the UK and top in Scotland
> > (National Student Survey 2012)
> >
> >
> > We invite research leaders and ambitious early career researchers to
> > join us in leading and driving research in key inter-disciplinary themes.
> > Please see www.hw.ac.uk/researchleaders for further information and
> > how to apply.
> >
> > Heriot-Watt University is a Scottish charity registered under charity
> > number SC000278.
> >
> >
> > *
> > * For searches and help try:
> > * http://www.stata.com/help.cgi?search
> > * http://www.stata.com/support/faqs/resources/statalist-faq/
> > * http://www.ats.ucla.edu/stat/stata/
>
>
>
> --
> Anat Tchetchik, PhD
> Department of Business Administration
> Guilford Glazer Faculty of Business and Management Ben-Gurion University of
> the Negev
> P.O.Box: 653
> Beer-Sheva, Israel, 84105
>
> E-mail: [email protected]
> Phone 972-(0)8-6479735
> Fax: 972-(0)8-6472920
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/faqs/resources/statalist-faq/
> * http://www.ats.ucla.edu/stat/stata/
-----
Sunday Times Scottish University of the Year 2011-2013
Top in the UK for student experience
Fourth university in the UK and top in Scotland (National Student Survey 2012)
We invite research leaders and ambitious early career researchers to
join us in leading and driving research in key inter-disciplinary themes.
Please see www.hw.ac.uk/researchleaders for further information and how
to apply.
Heriot-Watt University is a Scottish charity
registered under charity number SC000278.
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/faqs/resources/statalist-faq/
* http://www.ats.ucla.edu/stat/stata/