Question about AR test of xtabond

[Raymond Zhang]

Dear all,

I have a question about xtabond.We can use --estat abond--to do the AR test of the error terms.But I want to do AR test step by step.I mean ,firstly, get the error and then use --reg--command.
Here is the example.

webuse abdata,clear
 xtabond n w k, twostep
 estat abond
 
 Arellano-Bond test for zero autocorrelation in first-differenced errors
  +-----------------------+
  |Order |  z     Prob > z|
  |------+----------------|
  |   1  |-2.4955  0.0126 |
  |   2  |-.52671  0.5984 |
  +-----------------------+
   H0: no autocorrelation

  
 predict e,e
 reg e L.e L2.e
 
 
      Source |       SS           df       MS      Number of obs   =       611
-------------+----------------------------------   F(2, 608)       =   2845.84
       Model |  75.5173265         2  37.7586633   Prob > F        =    0.0000
    Residual |  8.06695695       608  .013268021   R-squared       =    0.9035
-------------+----------------------------------   Adj R-squared   =    0.9032
       Total |  83.5842835       610  .137023416   Root MSE        =    .11519

------------------------------------------------------------------------------
           e |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
           e |
         L1. |      0.681      0.042    16.35   0.000        0.599       0.763
         L2. |      0.254      0.041     6.22   0.000        0.174       0.334
             |
       _cons |     -0.013      0.005    -2.79   0.006       -0.023      -0.004
------------------------------------------------------------------------------

I find that result with --reg-- is different from the result of the command --estat abond--.
Anyone can give some suggestions?I want to know the calculation process behind --estat abond--.

Many thanks in advance.


Best regards.
Raymond

[Raymond Zhang]

I can get first-differenced errors by running "predict e, e", but what can I do in next step? Anyone can help me?@Nick Cox Carlo Lazzaro

[Nick Cox]

Sorry, but I have never used Arellano-Bond.

[Raymond Zhang]

Nick Cox Thank you for your reply.Another simple question,How can I control time trend in panel FE model? here are two methods, which method is right?

webuse grunfeld,clear
bys company:gen Trend=_n
xtreg invest mvalue kstock Trend,fe

xtreg invest mvalue kstock c.year,fe

Best regards.
Raymond

[Carlo Lazzaro]

Raymond:
the first advice is to switch form -xtreg- to -xtregar- or -xtgls- if you have a T>N panel dataset.

[Raymond Zhang]

@Carlo Lazzaro Thank you for your suggestion ,Carlo. But here ,I just want to know how to control time trend in FE model. If I want to add the time trend in the model ,how can I write the codes?

Best regards.
Raymond

[Carlo Lazzaro]

Raymond:
you cannot include -timevar- in -xtregar-.
In -xtreg, fe- (that was developed for N>T panel datasets) -timevar- is usually coded as a categorical predictor:

. use "https://www.stata-press.com/data/r16/nlswork.dta"
(National Longitudinal Survey.  Young Women 14-26 years of age in 1968)

. xtreg ln_wage i.year i.race, fe vce(cluster idcode)
note: 2.race omitted because of collinearity
note: 3.race omitted because of collinearity

Fixed-effects (within) regression               Number of obs     =     28,534
Group variable: idcode                          Number of groups  =      4,711

R-sq:                                           Obs per group:
     within  = 0.1059                                         min =          1
     between = 0.0788                                         avg =        6.1
     overall = 0.0721                                         max =         15

                                                F(14,4710)        =      74.45
corr(u_i, Xb)  = 0.0283                         Prob > F          =     0.0000

                             (Std. Err. adjusted for 4,711 clusters in idcode)
------------------------------------------------------------------------------
             |               Robust
     ln_wage |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
        year |
         69  |    .086922   .0104445     8.32   0.000     .0664458    .1073981
         70  |   .0730999   .0107089     6.83   0.000     .0521054    .0940943
         71  |   .1245316   .0113143    11.01   0.000     .1023503    .1467129
         72  |   .1361498   .0122887    11.08   0.000     .1120582    .1602414
         73  |   .1510882    .012356    12.23   0.000     .1268648    .1753117
         75  |   .1656841   .0124631    13.29   0.000     .1412506    .1901176
         77  |    .221329   .0129468    17.10   0.000     .1959472    .2467107
         78  |   .2570319    .013432    19.14   0.000     .2306989    .2833649
         80  |   .2650459   .0136801    19.37   0.000     .2382266    .2918652
         82  |   .2850749    .013733    20.76   0.000     .2581517    .3119981
         83  |   .3095487   .0141833    21.82   0.000     .2817427    .3373547
         85  |   .3601922    .013813    26.08   0.000     .3331123    .3872722
         87  |   .3771489   .0142489    26.47   0.000     .3492145    .4050833
         88  |   .4354505   .0153701    28.33   0.000     .4053178    .4655831
             |
        race |
      black  |          0  (omitted)
      other  |          0  (omitted)
             |
       _cons |   1.444088    .010313   140.03   0.000      1.42387    1.464307
-------------+----------------------------------------------------------------
     sigma_u |  .40806568
     sigma_e |  .30294584
         rho |  .64468296   (fraction of variance due to u_i)
------------------------------------------------------------------------------

.