Note that you report nine categories, and I don't think your CIs will be
plausible, given the number of obs and apparent weighting and survey design.
Plus your categories are suspect, since if they were 0/2 you would have 27
categories instead of nine. But mine is not to reason why. Using code from
jpitblado@stata.com,
di "cat1Rr cat2Rr cat3Rr (Lower Bound, Upper Bound)"
forvalues i=0/2 {
forvalues j=0/2 {
forvalues k=0/2 {
qui gen i`i'j`j'k`k'=(cat1Rr==`i' & cat2Rr==`j' & cat3Rr==`k')
qui su i`i'j`j'k`k'
if r(max)>0 & r(N)>0 {
qui svylogit i`i'j`j'k`k'
scalar lcb = invlogit(_b[_cons]-invttail(e(df_r),.025)*_se[_cons])
scalar ucb = invlogit(_b[_cons]+invttail(e(df_r),.025)*_se[_cons])
di " `i' `j' `k' ( " scalar(lcb) " , "
scalar(ucb) " )"
}
}
}
}
gives CIs that are constructed independently and cannot be used to
eyeball-test joint hypotheses about proportions. Caveat emptor.
-----Original Message-----
From: hk489@nyu.edu [mailto:hk489@nyu.edu]
Sent: Wednesday, September 01, 2004 5:35 PM
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: calculrating confidence Intervals in svyprop statements
Despite of some valuable comments of you, I couldn't solve the problem. At
the last reply, Jeff explained about how to get CI with "SVYMEAN"
statements. I need to calculate CIs with "SVYPROP" statements, however.
Below is the table explaining how I calculate 8 Est. Prop.
*
* 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/