st: sureg problem - comparison with nlsur

[Alex Olssen <alex.olssen@gmail.com>]

As a follow up to my last post I would like to further highlight the
problem by making a comparison between -sureg- and -nlsur-

Again I use examples from auto.dta

sysuse auto, clear
sureg (weight length = mpg)
nlsur (weight = {a0} + {a1}*mpg) (length = {b0} + {b1}*mpg)

A comparison shows that -nlsur- and -sur- produce the same results
when given the same linear model.

But now consider

sysuse auto, clear
sureg (weight length = mpg price)
nlsur (weight = {a0} + {a1}*mpg + {a2}*price) (length = {b0} +
{b1}*mpg + {b2}*price)

The two models are identical.  Stata estimates the first model but
complains about a singular covariance matrix for the second model.

Is this correct?  Can anybody please give suggestions as to what is
happening here?

Kind regards,

Alex Olssen
Motu Economic Research
*
*   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/