st: xtabond2 with large datasets

[Asdsa Sdjfkn <asquriot@yahoo.com>]

I'm trying to run xtabond2 models on a fairly large dataset - 8 million 
observations and ~350mb. The command I issue is of the form:

 xtabond2 yvar L.xvar, twostep robust noleveleq iv(L2.xvar)

which runs fine with person fixed effects, but when I use person-month fixed 
effects, I get the out of memory error:

          J():  3900  unable to allocate real <tmp>[1288703978,8]
         xtabond2_mata():     -  function returned error
                 <istmt>:     -  function returned error

The machine I'm running the command on has about 16gb memory. What can one do in 
this situation? I'm only using one instrumental variable and it's not in "GMM" 
form. Would ivregress, gmm fare better?

Thanks for any help.


      

*
*   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/