How do I obtain _b[] and _se[] at the same time from Bootstrap with panel data ?

[Ning Wei]

The structure of my data is a panel, under the command "xtreg", how can I derive _b[] and _se[] at the same time with 200 repetition?
code:
xtreg F`i'.g exp0bp rec0bp $bpnlxlist, fe vce(bootstrap)

gen bgexph`i' = _b[exp0bp]
gen bgrech`i' = _b[rec0bp]

gen segexph`i' = _se[exp0bp]
gen segrech`i' = _se[rec0bp]

[Joro Kolev]

Try:

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

.         . xtset idcode
       panel variable:  idcode (unbalanced)

. bootstrap _b _se, saving(TEMP, replace): xtreg ln_w age south, fe
(running xtreg on estimation sample)

Bootstrap replications (50)
----+--- 1 ---+--- 2 ---+--- 3 ---+--- 4 ---+--- 5 
..................................................    50

Bootstrap results                               Number of obs     =     28,502
                                                Replications      =         50

------------------------------------------------------------------------------
             |   Observed   Bootstrap                         Normal-based
             |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
b            |
         age |   .0181924   .0003541    51.38   0.000     .0174985    .0188864
       south |  -.0774963   .0153928    -5.03   0.000    -.1076655    -.047327
       _cons |   1.178256    .011039   106.74   0.000      1.15662    1.199892
-------------+----------------------------------------------------------------
se           |
         age |   .0003475   3.78e-06    91.97   0.000     .0003401    .0003549
       south |   .0112551   .0002045    55.02   0.000     .0108542     .011656
       _cons |   .0111487   .0001361    81.94   0.000      .010882    .0114154
------------------------------------------------------------------------------

. use TEMP
(bootstrap: xtreg)

. des

Contains data from TEMP.dta
  obs:            50                          bootstrap: xtreg
 vars:             6                          14 Sep 2020 09:32
 size:         1,200                          
----------------------------------------------------------------------------------------------------
              storage   display    value
variable name   type    format     label      variable label
----------------------------------------------------------------------------------------------------
_b_age          float   %9.0g                 _b[age]
_b_south        float   %9.0g                 _b[south]
_b_cons         float   %9.0g                 _b[_cons]
_se_age         float   %9.0g                 _se[age]
_se_south       float   %9.0g                 _se[south]
_se_cons        float   %9.0g                 _se[_cons]
----------------------------------------------------------------------------------------------------
Sorted by: 

. summ

    Variable |        Obs        Mean    Std. Dev.       Min        Max
-------------+---------------------------------------------------------
      _b_age |         50    .0182427    .0003541   .0174261   .0192864
    _b_south |         50    -.076995    .0153928  -.1069007  -.0536908
     _b_cons |         50    1.176522     .011039   1.143865   1.197643
     _se_age |         50    .0003396    3.78e-06   .0003308   .0003494
   _se_south |         50     .011177    .0002045   .0108174   .0117697
-------------+---------------------------------------------------------
    _se_cons |         50    .0109135    .0001361   .0106551   .0112061

.

[Ning Wei]

Thank you very much for this detailed answer!

[Joro Kolev]

You are welcome. It can also be passed as an option as you tried the first time around:

. xtreg ln_w age south, fe vce(bootstrap _b _se, saving(TEMP, replace))

Thank you very much for this detailed answer!

[Ning Wei]

That 's great! Thanks very much.

[Marco Tacchi]

Hello,

I'd like to follow up on this post. Would using

bootstrap, seed(1): xtreg ln_w age south, fe

be justified? In the above example, the standard errors are higher under

xtreg ln_w age south, fe vce(bootstrap, seed(1))

. bootstrap, seed(1): xtreg ln_w age south, fe
(running xtreg on estimation sample)

Bootstrap replications (50): .........10.........20.........30.........40.........50 done

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

R-squared:                                      Obs per group:
     Within  = 0.1044                                         min =          1
     Between = 0.1233                                         avg =        6.1
     Overall = 0.1062                                         max =         15

                                                Wald chi2(2)      =    1977.76
corr(u_i, Xb) = 0.0818                          Prob > chi2       =     0.0000

                                  (Replications based on clustering on idcode)
------------------------------------------------------------------------------
             |   Observed   Bootstrap                         Normal-based
     ln_wage | coefficient  std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
         age |   .0181924   .0004156    43.77   0.000     .0173779     .019007
       south |  -.0774963   .0131054    -5.91   0.000    -.1031824   -.0518101
       _cons |   1.178256   .0136785    86.14   0.000     1.151446    1.205065
-------------+----------------------------------------------------------------
     sigma_u |  .39998991
     sigma_e |  .30322383
         rho |  .63504833   (fraction of variance due to u_i)
------------------------------------------------------------------------------
F test that all u_i=0: F(4709, .) = 8.22                     Prob > F =      .

. xtreg ln_w age south, fe vce(bootstrap, seed(1))
(running xtreg on estimation sample)

Bootstrap replications (50): .........10.........20.........30.........40.........50 done

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

R-squared:                                      Obs per group:
     Within  = 0.1044                                         min =          1
     Between = 0.1233                                         avg =        6.1
     Overall = 0.1062                                         max =         15

                                                Wald chi2(2)      =    1188.57
corr(u_i, Xb) = 0.0818                          Prob > chi2       =     0.0000

                              (Replications based on 4,710 clusters in idcode)
------------------------------------------------------------------------------
             |   Observed   Bootstrap                         Normal-based
     ln_wage | coefficient  std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
         age |   .0181924   .0005292    34.38   0.000     .0171552    .0192297
       south |  -.0774963   .0174753    -4.43   0.000    -.1117473   -.0432453
       _cons |   1.178256   .0154093    76.46   0.000     1.148054    1.208457
-------------+----------------------------------------------------------------
     sigma_u |  .39998991
     sigma_e |  .30322383
         rho |  .63504833   (fraction of variance due to u_i)
------------------------------------------------------------------------------
F test that all u_i=0: F(4709, .) = 8.22                     Prob > F =      .

[Jeff Pitblado (StataCorp)]

Marco Tacchi is missing some necessary options to bootstrap for it to properly resample the panel data, please see the FAQ https://www.stata.com/support/faqs/s...th-panel-data/

[Felix Bittmann]

In other words, if a standard command does accept vce(bootstrap), there is not need to handle it manually. The VCE way is the easy and correct one. So you might want to go with

xtreg ln_w age south, fe vce(bootstrap, seed(1))