In most survey data, you will have pweights, not fweights, so you will
want to center specifying aweights (means are equiv using pweights and
aweights, but variance estimates are not).
For an artificial, contrived, overly simple example, try:
. ssc install center
. ssc install ivreg2
. gen pwt=pop/1000
. foreach v of varlist death marr div pop18 pop65 {
g r_`v'=`v'/pop
}
. qui tab regi, gen(rd)
. drop rd1
. bysort regi: center r_de r_pop18 r_pop65 r_di r_ma [aw=pwt], prefix(d)
. ivreg2 r_de r_pop18 r_pop65p (r_di =r_ma) rd* [pw=pwt], ffirst
. ivreg2 dr_de dr_pop18 dr_pop65p (dr_di =dr_ma) [pw=pwt], ffirst
On 5/9/05, Kit Baum <[email protected]> wrote:
<snip> To answer the first question, NO, these smell like classic
frequency weights to me. As for the second, you can do the weighting
in Jann's center command:
webuse census2
center medage [fw=pop]
In my understanding that will give you a demeaned, appropriately
weighted median age variable.
*
* 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/
