Example binary data

We calculate the sample size required for a future study involving four post-treatment visits required to detect a doubling in the odds ratio at the 5% significance level with 80% power. The difference in logit (log(odds ratio)) required isthenlog(2) — log(1) = 0.693. A correlation of p = 0.50 between the repeated measurements and a dispersion parameter of one will be assumed. We will assume that the expected rate of 'positives' is 0.4 and hence b = 0.4 x 0.6 = 0.24. The number of patients required per treatment group is n = 2(1.96 + 0.84)2 x (1.00/0.24) x (1 + 3 x 0.50)/(4 x 0.693)2 = 85.

Was this article helpful?

0 0

Post a comment