Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Creating a loop for bsample


From   Farasat Bokhari <farasat.bokhari@yahoo.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Creating a loop for bsample
Date   Sun, 26 Jul 2009 12:17:37 -0700 (PDT)

Ok.  Thanks, I will see what alternative is available to create subsample(s) where selection is based on the original weights.  Thanks,fb 



----- Original Message ----
> From: Austin Nichols <austinnichols@gmail.com>
> To: statalist@hsphsun2.harvard.edu
> Sent: Saturday, July 25, 2009 11:47:42 PM
> Subject: Re: st: Creating a loop for bsample
> 
> Farasat Bokhari:
> No--don't use the weight option in -bsample-.  If you want to do it
> the quick-n-dirty way, save the relevant subset of data, -use- and
> resample with -bsample- (unweighted resample) then do weighted
> estimation on each draw using the original sample weights (the N
> totals will be wrong in each resample, but this is likely not a big
> deal, depending on what you want)--or ask Stas Kolenikov (see also
> http://ideas.repec.org/p/boc/nsug08/19.html or find him in DC this
> week or next) about the right way (if you resample you are supposed to
> redo all the weights and poststratification etc. for each draw).
> 
> On Sat, Jul 25, 2009 at 4:22 PM, Farasat
> Bokhariwrote:
> > I think I have the looping and posting commands figured out and the results 
> looks fine.  However, I am now worried per your comment: I understand that 
> bsample changes the values of the weight variable to be 0,1,2, etc. to indicate 
> how many times this obs. is chosen in the sample BUT i thought it was doing the 
> selection based on the original values of the weights.  If this is not so the 
> clearly my code is not doing what I thought it was (else its ok).  My code is 
> given below (your thoughts would be appreciated).
> >
> > fb
> >
> > ------code below --------
> >
> >
> > set memory 99999k
> > set matsize 400
> > #delimit ;
> > cd I:\bokhari\fb_research\analysis\adhd\schoollaws\schoollaws_data;
> > use nschschlaws2003;
> >
> > #delimit;
> > svyset idnumr [pw=weight], strata(state);
> > gen myflag1 = 0;
> > replace myflag1 = 1 if sample1 == 1 & pubsch == 1;
> > gen wt2 = weight;
> >
> >
> > #delimit;
> > postfile mysim assistance rewards sanctions using simres2, replace ;
> >
> >  forvalues i = 1/100 {
> >  bsample 6700 if myflag1 == 1, weight(wt2)
> >  quietly: probit adhd assistance rewards sanctions gender age black othrace 
> hispanic family2 family3 family4 poverty1-poverty7 edu1 edu2 conachive1 
> conachive2 schcalls1 schcalls2 if wt2!=0  [fw=wt2]
> >  replace wt2 = weight
> >  local b1 = _b[assistance]
> >  local b2 = _b[rewards]
> >  local b3 = _b[sanctions]
> >  post mysim (`b1') (`b2') (`b3')
> >  display "Run Done:" `i'
> >  }
> > postclose mysim
> >
> > #delimit;
> > clear all;
> > #delimit;
> > use simres2;
> >
> > mean assistance rewards sanctions;
> > mean assistance rewards sanctions, level(90);
> >
> 
> *
> *   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/



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