Re: st: Non integer weights - problem

[Stas Kolenikov <skolenik@gmail.com>]

I second Nick -- there are few, if any, problems that would require
you to collapse your data with weights. What is the analytic task that
you are performing that requires this transformation? If you need to
compute some sort of totals, then you might be better off with

total whatever [pw=weights], over( grouping variables )

or better yet

svy : total whatever , over( grouping variables )

provided your data are -svyset- properly.

On Sun, Feb 5, 2012 at 5:12 PM, Raquel Guimaraes
<raquelrguimadem@gmail.com> wrote:
> Dear all,
>
> I had a problem using collapse and expanding my sample using frequency
> weights:
>
> . collapse (count) id [fw=wtper], by (granosest sex)
> may not use noninteger frequency weights
>
> However, I summarized the weight variable and they seem to be fine:
>
> . sum wtper
>
>    Variable |       Obs        Mean    Std. Dev.       Min        Max
> -------------+--------------------------------------------------------
>       wtper |   3971891    17.56003    8.800185          2     167.13
>
> In order to solve this, I tried to create new rounded values, but none was
> created:
>
> . replace wtper_rounded=0 if  wtper<0
> (0 real changes made)
>
> Also, I tried to recast, but the error persists.
>
> . recast int wtper
> wtper:  3932021 values would be changed; not changed
>
> . collapse (count) id [fw=wtper], by (granosest sex)
> may not use noninteger frequency weights



-- 
Stas Kolenikov, also found at http://stas.kolenikov.name
Small print: I use this email account for mailing lists only.

*
*   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/