st: Standard error of non-linear combination of 2 estimates: one from the population and another from a subpopulation

[Collado Diego <Diego.Collado@uantwerpen.be>]

Dear Statalist,

Is it possible to get the standard error of a non-linear combination of 2 non-independent estimates: one coming from the population and another from a subpopulation? For example, the quotient between the median of a group and the median of the total population.

The command nlcom allows getting standard errors after any Stata estimation command, however those commands (as far as I know) are always based on the same (sub)populations (i.e. same degrees of freedom), for example, epctile. 

One issue could be that the covariance has to be calculated among the same number of cases in each of the two variables. This can be done with correlate taking only the cases of the subpopulation. I imagine that later I can take the degrees of freedom of the subpopulation and apply a formula (or an approximation) to get the standard error of the quotient; however, this is not as efficient as, for example, using nlcom for estimates from the same (sub)populations.

Best wishes,

PS: the denominator of the quotient is never 0

Diego Collado
Universiteit Antwerpen
Herman Deleeck Centre for Social Policy
Stadscampus, Sint Jacobstraat 2
2000 Antwerpen
www.centrumvoorsociaalbeleid.be


*
*   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/