st: residuals change after reloading data

[Yogesh Uppal <stata.list.queries@gmail.com>]

Hi Statalisters,

There may be a real simple answer to it. But I can't seem to figure
out why. Suppose I run a regression of the following form to test for
serial correlation (as suggested by Wooldridge):

xi: ivreg2 y x1 x2 [x3=z1 z2 z3] i.id i.time, robust
predict e, resid
bys id: gen lag_e = e[_n-1]
xi: ivreg2 y x1 x2 [x3=z1 z2 z3] i.id i.time lag_e, robust

Every time I clear the data from memory and reload it and run the
regressions again, the residuals change and so do the results of the
serial correlation test. So, I am not sure which results to use.

I appreciate any advice regarding what I am missing here. Many thanks!

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