st: xtdpd - a question on "e(sample)"

[Johannes Geyer <JGeyer@diw.de>]

Dear Statalisters,

I run a dynamic GMM model with Stata. Now i want to create a table with 
observations per year 
used in the estimation. I use a second lag of the dependent variable. I 
thought, that valid 
observations would start from t=4 onwards when I use the model in first 
differences:

xtdpd L(0/2).Y X1 X2 ,  div(X1 X2) dgmmiv(Y) twostep

tab year if e(sample)

I get: 

[t = 1  .]
[t = 2  .]
t = 3   N_3
t = 4   N_4
t = 5   N_5
t = 6   N_6
t = 7   N_7
...

And I thougt, that I am estimating a model like the following:

(Y_i,t - Y_i,t-1) = A_1 (Y_i,t-1 - Y_i,t-2) + A_2  (Y_i,t-2 - Y_i,t-3) + 
....

Wouldn't the second lagged term be missing? Because (Y_i,t-3) is not 
defined for (t<=3) in my setting.
 Then I would expect to have no observations in the table for t=3. 

Any help is appreciated,

Johannes

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/