Kazi Mizanur Rahman wrote:
The baseline proportion of a an event was 0.9.
We assume that after providing an intervention in one group for a particular
time period we get the proportion as 0.5. We also have a comparison group
that receives no intervention. After that particular time period the
proportion becomes 0.8 in the comparison group. How to calculate the sample
size with 80% power and alpha .05 ?
--------------------------------------------------------------------------------
There are undoubtedly some analytical solutions to this power analysis, but
have you considered simulation? See A. H. Feiveson, Power by simulation.
_The Stata Journal_ 2(2):107-24, 2002. ( www.stata-journal.com/sj2-2.html for
an abstract.) Something analogous to the do-file below might be worth
considering. You'd need to take into account any assumptions regarding
within-subject correlation, if your study design calls for repeated
measurements.
Joseph Coveney
clear
set more off
set seed `=date("2005-01-06", "ymd")'
*
program define kazisim, rclass
replace res0 = uniform() < 0.9
replace res1 = uniform() < (0.8 - `1' * trt * 0.3)
logit res1 res0 trt, nolog
test trt
return scalar p = r(p)
end
* Power (under the alternative hypothesis)
forvalues samplesize = 50(10)100 {
display in smcl as result "`samplesize'"
clear
quietly set obs `samplesize'
generate byte trt = _n > _N / 2
quietly generate byte res0 = .
quietly generate byte res1 = .
simulate "kazisim 1" p = r(p), reps(1000)
generate byte pos = (p < 0.05)
summarize pos
}
* Test size (under the null hypothesis)
forvalues samplesize = 50(10)100 {
display in smcl as result "`samplesize'"
clear
quietly set obs `samplesize'
generate byte trt = _n > _N / 2
quietly generate byte res0 = .
quietly generate byte res1 = .
simulate "kazisim 0" p = r(p), reps(1000)
generate byte pos = (p < 0.05)
summarize pos
}
exit
*
* 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/
