How to save 1 specific xtreg beta coefficient to use in postestimations?

[Joao Sa Oliveira]

Dear Stata Community members,

I would like to ask how I can save one particular beta coefficient after running an xtreg with fixed effects (I need to save the coefficient, its standard error, P|z|, and 95% CI interval). Please note that my model has many variables but I only need to save one of the betas (the coefficient, its standard error, P|z|, and 95% CI interval).

My objective is to use that coefficient in different postestimations (done using nlcom) after running other xtreg’s in the meantime (my post-estimation involves using coefficients from different xtregs (1 from each xtreg).

I would also like to ask how I can extract that saved coefficient and “refer to it” in my postestimations (so that STATA understands which coefficient I am referring to). Is there any command similar to “global” which does such deed? I’ve tried several options. One member of our community kindly suggested the use of regress and suest: such command does achieve the goal but I have too many fixed effects (1500+ units, each of which having between 2-16 observations), which makes each post-estimation (nlcom) take about 1.5 hours. I need to do many estimations at this stage so this solution is not very efficient.

I’m struggling a lot with this so your vice would be much appreciated 😊

Kind regards,
Joao

[William Lisowski]

Perhaps this example will start you in a useful direction. Note that I am running Stata 16.1, if you have an earlier version some of the matrix commands may need to change in ways that I no longer remember.

webuse nlswork
xtset idcode
xtreg ln_wage age not_smsa south, fe
matrix list r(table)
scalar b = r(table)["b","age"]
scalar se = r(table)["se","age"]
display "coeficient estimate is " b " with standard error " se

. webuse nlswork
(National Longitudinal Survey.  Young Women 14-26 years of age in 1968)

. xtset idcode
       panel variable:  idcode (unbalanced)

. xtreg ln_wage age not_smsa south, fe

Fixed-effects (within) regression               Number of obs     =     28,502
Group variable: idcode                          Number of groups  =      4,710

R-sq:                                           Obs per group:
     within  = 0.1084                                         min =          1
     between = 0.1664                                         avg =        6.1
     overall = 0.1395                                         max =         15

                                                F(3,23789)        =     964.43
corr(u_i, Xb)  = 0.1246                         Prob > F          =     0.0000

------------------------------------------------------------------------------
     ln_wage |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         age |   .0182881   .0003469    52.72   0.000     .0176082    .0189679
    not_smsa |  -.1016459   .0098022   -10.37   0.000    -.1208588    -.082433
       south |  -.0683062   .0112649    -6.06   0.000    -.0903861   -.0462262
       _cons |   1.200435   .0113276   105.97   0.000     1.178232    1.222638
-------------+----------------------------------------------------------------
     sigma_u |  .39222024
     sigma_e |  .30254719
         rho |  .62695426   (fraction of variance due to u_i)
------------------------------------------------------------------------------
F test that all u_i=0: F(4709, 23789) = 7.73                 Prob > F = 0.0000

. matrix list r(table)

r(table)[9,4]
               age    not_smsa       south       _cons
     b   .01828805  -.10164593  -.06830616    1.200435
    se   .00034687   .00980219   .01126493    .0113276
     t   52.723746  -10.369713  -6.0636131   105.97435
pvalue           0   3.853e-25   1.351e-09           0
    ll   .01760818  -.12085885  -.09038613   1.1782322
    ul   .01896793    -.082433  -.04622618   1.2226378
    df       23789       23789       23789       23789
  crit   1.9600637   1.9600637   1.9600637   1.9600637
 eform           0           0           0           0

. scalar b = r(table)["b","age"]

. scalar se = r(table)["se","age"]

. display "coeficient estimate is " b " with standard error " se
coeficient estimate is .01828805 with standard error .00034687