Many thanks for all the response I received. With much help, I found the formula I need in "Statistical Methods in Medical Research" by P. Armitage, G. Berry, and J.N.S. Matthews.
n > {[z_alpha*sqrt(2p(1-p)) + z_beta*sqrt(p1(1-p1)+p2(1-p2))]/[p1-p2]}^2 + 2/abs(p1-p2)
where z_alpha = 1.96, z_beta = 0.8416, p1 = 0.5, p2 = 0.6, and p = (p1+p2)/2.
10% increase meant , in this case, an increase from 50% to 60%, and 407 is the number of patients required in each group. (In this case, the samples are taken in two different years, so it won't wreck the budget.)
Cheers, Jane.
-----Original Message-----
From: Cook Jane [mailto:[log in to unmask]]
Sent: 13 March 2003 10:46
To: [log in to unmask]
Subject: Sample size calculation
I've just joined up with a project that's already been going, and am trying
to duplicate a sample size calculation. The original proportion of the
population with a characteristic is 50%, and we want to detect a 10%
increase with 80% power at the 5% level of significance. I'm told we'd need
407 patients, but can't get this number myself. Could someone point me to a
formula or reference for sample size calculations.
Thanks, Jane.
---------------------------------
With Yahoo! Mail you can get a bigger mailbox -- choose a size that fits
your needs
---------------------------------
With Yahoo! Mail you can get a bigger mailbox -- choose a size that fits your needs
|