Statalist The Stata Listserver


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RE: xtivreg2, no constant in the outpuut?


From   "Schaffer, Mark E" <[email protected]>
To   <[email protected]>
Subject   st: RE: xtivreg2, no constant in the outpuut?
Date   Fri, 30 Jun 2006 16:01:33 +0100

Pierre,

> -----Original Message-----
> From: [email protected] 
> [mailto:[email protected]] On Behalf Of 
> Pierre Azoulay
> Sent: 30 June 2006 15:48
> To: [email protected]
> Subject: st: xtivreg2, no constant in the outpuut?
> 
> Hi, I am using xtivreg2 to estimate a fixed effects model 
> with "correct se's", ie, both heteroskedasticity and serial 
> correlation robust.
> 
> the data is a balanced panel of counties, 1991-1999.
> 
> We have year effects in there, but we cannot seem to get a 
> constant term in the regression. Are we doing something 
> wrong? I searched statalist but could not find anything...

-xtivreg2- doesn't report an estimated constant.  The reason, if I
recall correctly, is that it can cause problems when combining 2-step
GMM and cluster-robust (and maybe some other combinations - I would have
to check my notes).  The var-cov matrix of orthog conditions might not
be of full rank, etc. etc.

One alternative is to use -ivreg2- on data that have been demeaned and
then recentred, the way official -xtreg- and -xtivreg- do it.  Then you
would make a degrees-of-freedom adjustment by hand afterwards.  A bit
laborious, but do-able.

Another is to use official -xtivreg- but with cluster-robust standard
errors.  You've got a lot of cross-sections but only 9 time periods, and
the kernel approach (the -bw- option) is a bit of a stretch, since you
are relying on the time dimension to go off to infinity for the
asymptotics to kick in.  -cluster- gives you SEs that are robust to
arbitrary autotocorrelation and that rely only on the number of cross
sections going off to infinity.

Cheers,
Mark

Prof. Mark Schaffer
Director, CERT
Department of Economics
School of Management & Languages
Heriot-Watt University, Edinburgh EH14 4AS
tel +44-131-451-3494 / fax +44-131-451-3296
email: [email protected]
web: http://www.sml.hw.ac.uk/ecomes

> Thanks
> 
> Pierre
> 
> --------------------------------
> . tsset fips year;
>        panel variable:  fips, 1003 to 56021
>         time variable:  year, 1991 to 1999
> 
> . xi: xtivreg2    loghmoenrollees
> >                 logpop loginc logpop65 logpop15 lognwhite 
> logmdfp logmdnfp
> >                 i.year, robust fe bw(3);
> i.year            _Iyear_1991-1999    (naturally coded; 
> _Iyear_1991 omitted)
> 
> FIXED EFFECTS ESTIMATION
> ------------------------
> Number of groups =       560                    Obs per 
> group: min =         9
>                                                               
>  avg =       9.0
>                                                               
>  max =         9
> 
> OLS regression with robust standard errors
> ------------------------------------------
> Heteroskedasticity and autocorrelation-consistent statistics
>   kernel=Bartlett; bandwidth=3
>   time variable (t):  .
>   group variable (i): fips
> 
>                                                       Number 
> of obs =     5040
>                                                       F( 15,  
> 4465) =    46.51
>                                                       Prob > 
> F      =   0.0000
> Total (centered) SS     =  7772.500295                
> Centered R2   =   0.2324
> Total (uncentered) SS   =  7772.500295                
> Uncentered R2 =   0.2324
> Residual SS             =  5966.546306                Root 
> MSE      =    1.154
> 
> --------------------------------------------------------------
> ----------------
>              |               Robust
> loghmoenro~s |      Coef.   Std. Err.      z    P>|z|     
> [95% Conf. Interval]
> -------------+------------------------------------------------
> ----------
> -------------+------
>       logpop |   9.236974   2.213715     4.17   0.000     
> 4.898173    13.57578
>       loginc |  -1.267983   .7580286    -1.67   0.094    
> -2.753692    .2177256
>     logpop65 |  -2.955111   .7997608    -3.69   0.000    
> -4.522614   -1.387609
>     logpop15 |  -7.820074   1.613504    -4.85   0.000    
> -10.98248   -4.657665
>    lognwhite |  -.9190482   .9299859    -0.99   0.323    
> -2.741787    .9036906
>      logmdfp |   1.386816   .3748129     3.70   0.000     
> .6521959    2.121436
>     logmdnfp |   -.121715   .1130446    -1.08   0.282    
> -.3432785    .0998484
>  _Iyear_1992 |   .3863159   .0778425     4.96   0.000     
> .2337475    .5388843
>  _Iyear_1993 |   .8241919   .1151342     7.16   0.000     
> .5985329    1.049851
>  _Iyear_1994 |   1.148836   .1434591     8.01   0.000     
> .8676612     1.43001
>  _Iyear_1995 |   1.460995   .1694574     8.62   0.000     
> 1.128865    1.793126
>  _Iyear_1996 |   1.684981   .2008649     8.39   0.000     
> 1.291293    2.078669
>  _Iyear_1997 |   1.821555   .2342937     7.77   0.000     
> 1.362347    2.280762
>  _Iyear_1998 |   1.902581   .2735162     6.96   0.000     
> 1.366499    2.438663
>  _Iyear_1999 |   1.874201   .3029448     6.19   0.000      
> 1.28044    2.467962
> --------------------------------------------------------------
> ----------------
> 
> 
> --------------------------------------------------------------
> --------------------
> Pierre Azoulay
> Associate Professor of Management
> Columbia University                           Phone: (212) 854-9684
> Graduate School of Business                Fax:   Don't send any
> 3022 Broadway, Uris Hall 704
> New York, NY 10023                        
> http://www.columbia.edu/~pa2009/
> *
> *   For searches and help try:
> *   http://www.stata.com/support/faqs/res/findit.html
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
> 
> 

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index