st: system of 3 panel equations (dependents and lags on the RHS)

["nicola dott². ing. baldini" <n.baldini@unibo.it>]

I have three main variables (panels N*T). The most complete and complex model in my nightmares consists in regressing each one of three on the remaining two, on some lags of all three, and on some exogenous. Unfortunately, I cannot take advantage of the wonderful -gmm- command, as the later version of Stata to which I have access to date is 9.2
(1) Is there anything else I can use to estimate the above?
(2) I suppose that, if I reduce to the simplest model by ignoring the panel nature of the data and the lags, I may use -reg3-. Can use dummies for time/individuals as exogenous variables?
(3) Is there any command available to estimate something not as complex as the most difficult and not as easy as the simplest model? E.g. is it of any help if need to regress each variable on only the lags of the other two and on some exogenous?
(4) can I first difference by hand and then use -reg3-, or does first differencing require some error correction techniques which are not available in -reg3- (can't I correct via -bootstrap- or -jackknife-)? 
(5) To tell the truth, I may observe one variable for more time, let's say 5 years before time zero until time T, so that it is N*(T+5). Is it useful? 
(6) It's not very attractive, but now suppose I can sum over individuals, so to reduce the panels N*T to a time series: would you change your previous answers? 
Please note that my programming ability and my knowledge of Mata are below zero.

Many thanks in advance for your comments,
Nicola

*
*   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/