Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

RE: st: -estout- and ebsd


From   "Carter Rees" <carterrees@gmail.com>
To   <statalist@hsphsun2.harvard.edu>
Subject   RE: st: -estout- and ebsd
Date   Tue, 25 Sep 2007 10:12:00 -0400

Ben,

I believe that I have the issued worked out.  Thanks for the help.

Following the suggestions given in
http://www.stata.com/support/faqs/stat/supweight.html
I ran my model (svy, subpop(maleage): nbreg) and pulled the standard
deviations of my independent variables using both the -estadd- and
-listcoef- methods.  As mentioned before they are not the same.  From here I
wanted to see if I could use the above mentioned FAQ to reproduce the
-estadd- and -listcoef- sd's.  

summarize persfight [aw=gs] if maleage == 1 & e(sample) produced the same
sd's as -estadd- with an n = 3353.

summarize persfight [aw=gs] if e(sample) produced the same sd's as
-listcoef- with an n = 8619.

The difference (as you mentioned earlier) is in the use of the
subpopulation.  -listcoef- computes the sd's across all cases (in this
instance males and females ages 12 to 18 yrs) which is not what I wanted
here. -estadd- limits the calculation only to males in the appropriate age.

Thanks again.

Carter 

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Ben Jann
Sent: Tuesday, September 25, 2007 4:40 AM
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: -estout- and ebsd

Carter wrote:
> Your response seems to indicate that my survey
> design is correctly taken into account by -estadd- ebsd option (correct?).

Well, I'm not sure. It takes into account the weights and the
-subpop()- option, but -estadd ebsd- just uses plain -summarize- and
no adjustment is made for degrees of freedom or so, which I assume
would have to if there are strata and/or clusters. What would be the
"correct" formula for the SD in this case? Is there a Stata command
that computes it?
ben

On 9/25/07, Carter Rees <carterrees@gmail.com> wrote:
> Ben,
>
> The command I am using is svy subpop(): nbreg, so yes the subpop option is
> specified for my model.  Your response seems to indicate that my survey
> design is correctly taken into account by -estadd- ebsd option (correct?).
> Thus, maybe -listcoef- isn't recognizing the survey design.  (I had no
real
> reason to assume that -listcoef- was correct vs. -estadd-).  I am going to
> compare results between the two using your code below and will post a
brief
> summary of findings.
>
> Carter
>
> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu
> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Ben Jann
> Sent: Monday, September 24, 2007 3:30 PM
> To: statalist@hsphsun2.harvard.edu
> Subject: Re: st: -estout- and ebsd
>
> I don't know about -listcoef-, but -estadd ebsd- computes the sd using
> -summarize- on the observations identified by e(sample) (and e(subpop)
> if defined), and applying the appropriate weights. Did you use the
> -subpop()- option in your models? This might explain the difference.
>
> If you want tabulate results from listcoef, then do something like the
> following:
>
> . nbreg ...
> . listcoef, matrix  //-matrix- saves results in r(); type -return
> list- for details
> . matrix b_facts = r(b_facts)
> . estadd matrix b_facts = b_facts
> . estout ., cell(b b_facts) style(smcl)
>
> ben
>
> PS: A set of -estadd- commands to support the -spost- package is in
> preparation.
>
> On 9/24/07, Carter Rees <carterrees@gmail.com> wrote:
> > Statalist,
> >
> > Windows XP, Stata 10 SE.
> >
> > I am using svy: nbreg to run a series of regressions and then using
> -estout-
> > to format my regression tables.  Instead of the raw b's my tables
display
> > the standardized factor change coefficients via the -estadd- ebsd
option.
> >
> > All is well and good except the standardized factor change coefficients
> > don't precisely match those displayed if I check them using -listcoef-,
> > help.  My assumption is that -listcoef- takes into account the survey
> design
> > when calculating the sd of the independent variables while -estadd- ebsd
> may
> > not.  If this is the case, how can I include the standardized factor
> change
> > coef's as computed by -listcoef- in my -estout- generated table?
> >
> > Much appreciated.
> >
> > Carter
> >
> >
> > *
> > *   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/
>
> *
> *   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/

*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index