st: xtreg reports a range of obs per group that I don't understand

["M. Douglas Berg" <eco_mdb@shsu.edu>]

Hello,

I am using xtreg to estimate the relationship between market price "p" 
and a computed intrinsic value "iv". The data is organized by company 
identifier "permno" and year "yr". The variable "piv" is the ratio of 
"p" divided by "iv". I create a variable called group_size to represent 
the number of observations per permno. I then drop observations where 
group_size is less than 4. In other words I want each company in my 
dataset to have at least 4 years worth of data. When I do
   xtreg p iv if  piv>=1 & piv<50, re
the output reports that the minimum number of observations per group is 
1. I was expecting that this number should be 4 or greater. Could 
someone please explain?
I include part of the log file below.

Thanks for any help you may provide.

Doug Berg

-------------------------------------------------------------------------------------------------
use astockdata.dta, clear

.
. tsset permno yr
      panel variable:  permno, 10001 to 93113
       time variable:  yr, 1968 to 2002, but with gaps
. /*  Create a group size variable. Drop permno groups where there are 
fewer than 4 observat
> ions. */
. gen group_size=0

.
. by permno: replace group_size=_N
(51463 real changes made)

.
. drop if group_size < 4
(4347 observations deleted)

. /* Performing the Hausman test */
.
. xtreg p iv if piv>=1 & piv<50, re

Random-effects GLS regression                   Number of obs      =     
34147
Group variable (i): permno                      Number of groups   
=      4074

R-sq:  within  = 0.1542                         Obs per group: min 
=         1
      between = 0.4861                                        avg 
=       8.4
      overall = 0.3070                                        max 
=        30

Random effects u_i ~ Gaussian                   Wald chi2(1)       =   
8031.61
corr(u_i, X)       = 0 (assumed)                Prob > chi2        =    
0.0000

------------------------------------------------------------------------------
          p |      Coef.   Std. Err.      z    P>|z|     [95% Conf. 
Interval]
-------------+----------------------------------------------------------------
         iv |   .8689448    .009696    89.62   0.000      .849941    
.8879485
      _cons |   17.02923   .2104142    80.93   0.000     16.61682    
17.44163
-------------+----------------------------------------------------------------
    sigma_u |  10.679944
    sigma_e |  14.693601
        rho |  .34567866   (fraction of variance due to u_i)
------------------------------------------------------------------------------

. estimates store random_effects

. xtreg p iv if piv>=1 & piv<50, fe

Fixed-effects (within) regression               Number of obs      =     
34147
Group variable (i): permno                      Number of groups   
=      4074

R-sq:  within  = 0.1542                         Obs per group: min 
=         1
      between = 0.4861                                        avg 
=       8.4
      overall = 0.3070                                        max 
=        30

                                               F(1,30072)         =   
5480.55
corr(u_i, Xb)  = 0.3626                         Prob > F           =    
0.0000

------------------------------------------------------------------------------
          p |      Coef.   Std. Err.      t    P>|t|     [95% Conf. 
Interval]
-------------+----------------------------------------------------------------
         iv |   .7437109    .010046    74.03   0.000     .7240203    
.7634014
      _cons |    20.6917   .1151006   179.77   0.000      20.4661    
20.91731
-------------+----------------------------------------------------------------
    sigma_u |  14.504545
    sigma_e |  14.693601
        rho |  .49352532   (fraction of variance due to u_i)
------------------------------------------------------------------------------
F test that all u_i=0:     F(4073, 30072) =     8.11         Prob > F = 
0.0000

. hausman . random_effects

                ---- Coefficients ----
            |      (b)          (B)            (b-B)     
sqrt(diag(V_b-V_B))
            |       .       random_eff~s    Difference          S.E.
-------------+----------------------------------------------------------------
         iv |    .7437109     .8689448       -.1252339        .0026287
------------------------------------------------------------------------------
                          b = consistent under Ho and Ha; obtained from 
xtreg
           B = inconsistent under Ha, efficient under Ho; obtained from 
xtreg

   Test:  Ho:  difference in coefficients not systematic

                 chi2(1) = (b-B)'[(V_b-V_B)^(-1)](b-B)
                         =     2269.71
               Prob>chi2 =      0.0000




*
*   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/