Bookmark and Share

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]

st: RE: e(sample) after xtivreg2


From   "Schaffer, Mark E" <[email protected]>
To   <[email protected]>
Subject   st: RE: e(sample) after xtivreg2
Date   Tue, 12 Apr 2011 14:06:19 +0100

Chris,

To answer your specific question below:

> Can someone explain what is going on here? Is xtivreg2 actually using
> the 172 observations or not?

-xtivreg2- includes the 172 singletons in e(sample).  But of course the
FE (or FD) transformation effectively wipes these out, because there's
no within-group variation when there's only one observation in the
group.  -xtivreg2- warns about this.

-xtreg- also does not - cannot! - use singletons to estimate the slope
coefficients with the FE estimator.  However, -xtreg- reports a constant
term, and if I'm not mistaken, it *does* use the singletons for
estimating that.

I hope this clarifies things.

Cheers,
Mark

> -----Original Message-----
> From: [email protected] 
> [mailto:[email protected]] On Behalf Of 
> Chris Parker
> Sent: Tuesday, April 12, 2011 12:52 PM
> To: [email protected]
> Subject: st: e(sample) after xtivreg2
> 
> Hello statalist,
> 
> I am using xtivreg2 from ssc to perform two-way error clustering.
> However I would also like to look at the iid errors as well. I've
> noticed that xtreg is faster that xtivreg2 so I wanted to calculate
> the iid errors using xtreg and then use xtivreg2 only to calculate the
> two-way clustered errors.
> 
> . xtreg highprice volume, fe
> 
> Fixed-effects (within) regression               Number of obs 
>      =   1845327
> Group variable: cropstat~tid                    Number of 
> groups   =      8825
> 
> R-sq:  within  = 0.0005                         Obs per 
> group: min =         1
>        between = 0.0006                                       
>  avg =     209.1
>        overall = 0.0000                                       
>  max =       873
> 
>                                                 F(1,1836501)  
>      =    852.88
> corr(u_i, Xb)  = -0.0140                        Prob > F      
>      =    0.0000
> 
> --------------------------------------------------------------
> ----------------
>    highprice |      Coef.   Std. Err.      t    P>|t|     
> [95% Conf. Interval]
> -------------+------------------------------------------------
> ----------------
>       volume |  -.0037862   .0001296   -29.20   0.000    
> -.0040403   -.0035321
>        _cons |   1595.633   .4527128  3524.60   0.000     
> 1594.745     1596.52
> -------------+------------------------------------------------
> ----------------
>      sigma_u |  3092.7207
>      sigma_e |   602.9378
>          rho |  .96338464   (fraction of variance due to u_i)
> --------------------------------------------------------------
> ----------------
> F test that all u_i=0:     F(8824, 1836501) =  1758.26       
> Prob > F = 0.0000
> r; t=28.67 12:32:57
> 
> Which, on my four core license takes 28.67 seconds and uses 1,845,327
> observations. Now when I run the same regression using xtivreg2 I am
> issued a warning that singleton groups are detected and 172
> observations will not be used:
> 
> 
> . xtivreg2 highprice volume, fe
> Warning - singleton groups detected.  172 observation(s) not used.
> 
> FIXED EFFECTS ESTIMATION
> ------------------------
> Number of groups =      8653                    Obs per 
> group: min =         2
>                                                               
>  avg =     213.2
>                                                               
>  max =       873
> 
> OLS estimation
> --------------
> 
> Estimates efficient for homoskedasticity only
> Statistics consistent for homoskedasticity only
> 
>                                                       Number 
> of obs =  1845155
>                                                       F(  
> 1,1836501) =   852.88
>                                                       Prob > 
> F      =   0.0000
> Total (centered) SS     =  6.67941e+11                
> Centered R2   =   0.0005
> Total (uncentered) SS   =  6.67941e+11                
> Uncentered R2 =   0.0005
> Residual SS             =  6.67631e+11                Root 
> MSE      =    602.9
> 
> --------------------------------------------------------------
> ----------------
>    highprice |      Coef.   Std. Err.      z    P>|z|     
> [95% Conf. Interval]
> -------------+------------------------------------------------
> ----------------
>       volume |  -.0037862   .0001296   -29.20   0.000    
> -.0040403   -.0035321
> --------------------------------------------------------------
> ----------------
> Included instruments: volume
> --------------------------------------------------------------
> ----------------
> r; t=50.46 12:34:19
> 
> Just a note that this does in fact take longer, 50.46 seconds. Also,
> the 172 observations are not included in the regression as noted by
> number of obs. I thought it would be best to look at the xtreg
> regression without the 172 observations. So I attempted to run the
> xtreg regression with the sample from xtivreg2 only to find that all
> of the observations are included in the sample from xtivreg2.
> 
> . gen mysample = e(sample)
> r; t=0.20 12:35:04
> 
> . tab mysample
> 
>    mysample |      Freq.     Percent        Cum.
> ------------+-----------------------------------
>           1 |  1,845,327      100.00      100.00
> ------------+-----------------------------------
>       Total |  1,845,327      100.00
> r; t=0.56 12:35:07
> 
> Can someone explain what is going on here? Is xtivreg2 actually using
> the 172 observations or not?
> 
> My plan is to drop the singleton groups. However, I am running
> regressions with different groups of independent variables with
> missing observations. This could potentially create singleton groups
> again. Is there a way to make xtreg ignore the singleton groups
> automatically as well?
> 
> Additionally, does anyone know of a command that will let me calculate
> the two-way clustered errors after xtreg? I don't really see the point
> in running the same regression again just to cluster the errors
> differently but short of coding the two-way clustered errors myself I
> don't know how else to do this.
> 
> Chris
> 
> 
> Chris Parker
> 
> ________________________________
> 
> PhD Candidate | Management Science & Operations
> London Business School | Regent's Park | London NW1 4SA | 
> United Kingdom
> Direct line +44 (0)20 7000 8816 | Email [email protected]
> *
> *   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 charity
registered 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/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index