st: latent variable in nl and regression with error?

["Hoffman, George" <ghoffman@mcw.edu>]

Dear statalisters:

I have an xt dataset with multiple measures on subjects. (about 250 subjects, about 50 measures per subject).

The measurement device had an offset of either 10 or 0. this offset was constant across observations within subjects but it was not known which subject had which offset.

There is a theoretical relationship at each observation:
	y = m1*x1 + m2*x2 + b + offset 
which would be a simple linear regression if offset had been recorded for each subject.

Since offset was not recorded, I wish to fit the best model with offset of either 10 or 0 for each subject. 

I don't think nl can do this with different {offset} for each subject. I would like to come up with the best fit, and to fit a dichotomous offset for each subject.

Thanks in advance for orienting help. 
George


*
*   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/