RE: st: capturing estimated means and standard errors into a new data set

["Scholes, Shaun" <s.scholes@ucl.ac.uk>]

For survey data I highly recommend the -parmby- programme. I think I remember struggling with putting results from -svy- commands into matrices which thankfully stopped as soon I became aware of Roger Newson's programme.
The help file gives many useful examples including its use with survey data. I hope this reference is sufficient (SJ3-3: st0043) or maybe just type -search parmby-
Best wishes
Shaun


-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Kevin Geraghty
Sent: 16 May 2012 01:12
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: capturing estimated means and standard errors into a new data set

A good idea, and I'm glad to know about -statsby-. But unfortunately -statsby- appears not to be compatible with svy: prefix commands.  I expect it would be the best approach for non-complex-sample data.


----- Original Message -----
From: "Nick Cox" <njcoxstata@gmail.com>
To: statalist@hsphsun2.harvard.edu
Sent: Tuesday, May 15, 2012 3:56:46 PM
Subject: Re: st: capturing estimated means and standard errors into a new data set

I've not tried to understand the details here, but it sounds like a case for -statsby-.

Nick

On Tue, May 15, 2012 at 11:53 PM, Kevin Geraghty <kevin@ecotope.com> wrote:
> I am sure many have confronted this problem:
>
> I have a variable with some partition of subgroups defined , and want to capture the mean of that variable and accompanying standard error, for each value of the subgroup partition, into a stata data set.
>
> I have written code to confront this problem pulling data out of the returned  matrices and dumping it into a temporary dset, but it is frankly pretty ugly and I am hoping someone can show me a more economical way to do this.
>
> Here's what I have. note that "plat" is the name of the partition variable which defines subgroups. I am using the complex survey version of the mean command, but I believe the problem is identical with the standard version of the mean command:
>
>    tempfile outfile
>    local testvar <name of variable of interest>
>
>    svy: mean `testvar', over(plat) level(90)
>    preserve
>        clear
>        matrix fu = e(b)'
>        local vplatnames: rowfullnames fu
>        local plats: subinstr local vplatnames "`testvar':" "", all
>        svmat fu, names(col)
>        rename y1 `testvar'_mean
>        format `testvar'_mean %9.3f
>        gen int plat=.
>        local i 0
>        foreach plat of local plats {
>            replace plat = `plat' in `++i'
>        }
>        save `outfile', replace
>
>        clear
>        matrix var = vecdiag(e(V))'
>        local vplatnames: rowfullnames var
>        local plats: subinstr local vplatnames "`testvar':" "", all
>        svmat var, names(col)
>        replace r1 = sqrt(r1)
>        rename r1 `testvar'_sd
>        format `testvar'_sd %9.6f
>        gen int plat=.
>        local i 0
>        foreach plat of local plats {
>            replace plat = `plat' in `++i'
>        }
>        merge 1:1 plat using `outfile', nogenerate
>        save `outfile', replace
>    restore
>

*
*   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/

*
*   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/


*
*   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/