Hi,
I try to calculate the sample size for the following design:
- endpoint: response rate
- response rate placebo: 15%
- relative risk: 3
- 9 arms; 8 arms with Novum are compared versus Placebo
- alpha(overall): 5%
- Power: 80%
- it is sufficient, if 1 arm is better than placebo
My sample size program allows only 2 arms, so I made a Sidak alpha correction (for the 8 comparisons), so alpha goes down from 5% to 0.639%. I come up with n=60 patients per arm.
The problem now is: using a simulation I realized that the power increased to 98%.
So do I have to make the Sidak correction as well for the power? I tried this with my simulation but this seems not to work (if I use a Sidak corrected power of 18.2%, I get n=17 per arm, but with this my simulation shows less than 40% power).
Any help is appreciated, as well literature recommendation.
Thanks, Chr.
You may leave the list at any time by sending the command
SIGNOFF allstat
to [log in to unmask], leaving the subject line blank.
|