Hello,
I have a panel of large N and small T. Unbalanced and with time gaps. I hope to instrument the variables without including a lagged dependent variable. I use xtdpdgmm to perform the estimation. My codes are as follows:
Code:
xtdpdgmm idx_study avgPAI propcategory1-propcategory10 ///
parent_age cd_gender sleep wc4_b_1 edu_savings parenting, ///
model(fod) ///
gmm(avgPAI, bodev lag(1 .) ) ///
gmm(avgPAI, lag(1 1) diff model(level)) ///
gmm(parenting , bodev lag(1 .) ) ///
gmm(parenting , lag(1 1) diff model(level)) ///
gmm(propcategory1-propcategory10 wc4_b_1 parent_age sleep , bodev lag(0 .)) ///
gmm(propcategory1-propcategory10 wc4_b_1 parent_age sleep , lag(0 0) diff model(level)) ///
iv(cd_gender edu_savings, model(level)) ///
teffects two vce(r)
But when I ran the AR test, it returns missing values.
Code:
Arellano-Bond test for autocorrelation of the first-differenced residuals
H0: no autocorrelation of order 1:     z =         .   Prob > |z|  =         .
H0: no autocorrelation of order 2:     z =         .   Prob > |z|  =         .
I checked the previous posts and see that might be resulted from insufficient observations per group, so I add an if condition to restrict the sample to those who are observed for at least 3 times:

Code:
xtdpdgmm idx_study avgPAI propcategory1-propcategory10 ///
parent_age cd_gender sleep wc4_b_1 edu_savings parenting if (dynamic>2), ///
model(fod) ///
gmm(avgPAI, bodev lag(1 .) ) ///
gmm(avgPAI, lag(1 1) diff model(level)) ///
gmm(parenting , bodev lag(1 .) ) ///
gmm(parenting , lag(1 1) diff model(level)) ///
gmm(propcategory1-propcategory10 wc4_b_1 parent_age sleep , bodev lag(0 .)) ///
gmm(propcategory1-propcategory10 wc4_b_1 parent_age sleep , lag(0 0) diff model(level)) ///
iv(cd_gender edu_savings, model(level)) ///
teffects two vce(r)
The results are:

Code:
Group variable: num_child                    Number of obs         =      2723
Time variable: year                          Number of groups      =      2014

Moment conditions:     linear =      43      Obs per group:    min =         1
                    nonlinear =       0                        avg =  1.352036
                        total =      43                        max =         3

                            (Std. Err. adjusted for 2,014 clusters in num_child)
--------------------------------------------------------------------------------
               |              WC-Robust
    idxa_study |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
---------------+----------------------------------------------------------------
        avgPAI |   .5021869   .2527276     1.99   0.047       .00685    .9975238
 propcategory1 |   .1190317   .0801088     1.49   0.137    -.0379786     .276042
 propcategory2 |  -.5059096   .5617323    -0.90   0.368    -1.606885    .5950656
 propcategory3 |   .0418371   .4149987     0.10   0.920    -.7715455    .8552197
 propcategory4 |  -.0130479   .0470279    -0.28   0.781    -.1052209    .0791252
 propcategory5 |   .3501446    .280725     1.25   0.212    -.2000663    .9003556
 propcategory6 |   .5062244   .5492129     0.92   0.357    -.5702132    1.582662
 propcategory7 |  -.1510797   .1761881    -0.86   0.391     -.496402    .1942427
 propcategory8 |  -.6715729   .4368173    -1.54   0.124    -1.527719    .1845733
 propcategory9 |  -.7713343    .750685    -1.03   0.304     -2.24265    .6999812
propcategory10 |   .3808067   2.666237     0.14   0.886    -4.844922    5.606535
    parent_age |   .0070643   .0059831     1.18   0.238    -.0046623    .0187908
     cd_gender |  -.0600884   .0133498    -4.50   0.000    -.0862535   -.0339234
         sleep |   .0145435   .0332031     0.44   0.661    -.0505335    .0796204
       wc4_b_1 |  -.0149009   .0174536    -0.85   0.393    -.0491094    .0193076
   edu_savings |   .0410117   .0225825     1.82   0.069    -.0032492    .0852725
     parenting |   .0241621   .0339032     0.71   0.476    -.0422871    .0906112
               |
          year |
         2013  |  -.3513673   .1278175    -2.75   0.006     -.601885   -.1008497
         2014  |    -.02285   .0902107    -0.25   0.800    -.1996597    .1539597
         2015  |  -.1024287    .068121    -1.50   0.133    -.2359435    .0310861
         2016  |  -.1057155   .0797179    -1.33   0.185    -.2619598    .0505288
         2017  |          0  (empty)
         2018  |  -.0625284   .0786462    -0.80   0.427    -.2166722    .0916153
               |
         _cons |  -.0199318   .3966447    -0.05   0.960    -.7973411    .7574775
--------------------------------------------------------------------------------
Instruments corresponding to the linear moment conditions:
 5, model(fodev):
   2014:B.propcategory1 2014:B.propcategory3 2014:B.propcategory5
   2014:B.propcategory7 2014:B.propcategory9 2014:B.wc4_b_1 2014:B.sleep
   2014:B.propcategory2 2014:B.propcategory2 2014:B.propcategory4
   2014:B.propcategory6 2014:B.propcategory8 2012:L1.B.propcategory8
   2013:L1.B.propcategory10 2014:B.parent_age
 6, model(level):
   2013:D.parent_age 2014:D.cd_gender 2014:D.edu_savings 2014:D.2013bn.year
   2013:D.2014.year 2014:D.2015.year 2014:D.2016.year 2013:D.2018.year
   2014:D._cons 2014:D._cons 2014:D._cons 2014:D._cons 2014:D._cons
   2013:D._cons 2014:D._cons 2014:D._cons 2013:D._cons 2014:D._cons
   2013:D._cons 2014:D._cons
 7, model(level):
   _cons _cons
 8, model(level):
   _cons _cons _cons _cons _cons
 9, model(level):
   _cons
However, although the number of Obs. becomes much smaller than without the if condition, the average Obs. per groups just increased from 1.27 to 1.35. The AR tests are still missing.

My dataset looks like:

Can anyone help me? And does it sill make sense to estimate the serial correlation if no lagged dependent variable is added to the model?