You are correct, this is implausible. One of my programmers was using
a svy: tab command and taking the cell proportions to calculate row
proportions. The other was using the svy: proportion command to get
the row proportions directly. The rounding error in converting cell
proportions to row proportions accounted for the difference...
When I looked more closely at what they had done, I suggested that
they should have used the "row" option in the svy: tab command to
obtain row proportions instead of taking cell proportions and manually
calculating row proportions (doh!).
On Dec 21, 2007 10:03 AM, Austin Nichols <[email protected]> wrote:
> Lawrence Hanser <[email protected]>:
> That seems implausible. Can you provide commands and output? Can you
> duplicate the problem on a publicly available dataset, e.g. by
> starting with
> webuse nhanes2
> ?
>
> On Dec 21, 2007 12:07 PM, Lawrence Hanser <[email protected]> wrote:
> > Dear Colleagues:
> >
> > We're getting slightly different estimates (up to .04) using the svy:
> > proportions command vs. the svy: tab command (weighted with p
> > weights). Does anyone know why? Is one of these methods more
> > accurate for computing weighted proportions?
>
> *
> * 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/
>
*
* 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/
