Brent Fulton <fultonb@berkeley.edu> has a few questions regarding the
-subpop()- option on -svy:
> With Stata 9, I am using survey regression with the subpop command e.g.,
>
> svy, subpop(var1): reg dep1 ind1
>
> The regression result reports "Number of obs" and "Subpop. no. of obs".
> Based on running a few scenarios, the "Number of obs" includes all
> observations that have non-missing values for the regression variables,
> including the weights. This is as expected; however, the "Subpop. no. of
> obs" seems to include all observations that have a "1" for the subpop,
> including those observations that have missing values for the regression
> variables.
>
> An actual example is below where I would have thought "Subpop. no. of obs"
> would have been 5329, not 5436. (This is a simple example. I am estimating
> several models with different sets of variables and am only able to find
> out how many observations Stata is really using by running an extra
> command such as: count if !missing(var1, var2, etc.).)
>
> Lastly, I know if I am using the Taylor series approach to estimate
> parameter standard errors within a subpopulation, subpop is required
> (i.e., this is wrong: svy: reg dep1 ind1 if female==1). However, if I am
> using replicate weights, is subpop required?
-svy jackknife- was failing to update the variable that identifies the
supopulation sample when the prefixed estimation command drops observations
due to missing values (or other reasons).
-svy linearized- and -svy brr- do not have this problem.
We should have -svy jackknife- fixed by the next ado-file update.
--Jeff
jpitblado@stata.com
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/