3
$\begingroup$

I need to make a power analysis calculation to estimate the minimum sample size for a relative risk statistic (ratio of two proportions) for unmatched data - two independent samples.

Can I use the power calculations in the R pwr package for testing the difference between two proportions? Or will this return inaccurate sample size estimates?

EDIT: I see there's a power.cmh.test option in the samplesizeCMH package that allows you to find power and sample size for Cochran-Mantel-Haenszel tests. Can I apply this to any relative risk statistic, even if it's not a case-control study?

$\endgroup$
6
$\begingroup$

The relative risk, $RR$, is simply $RR = \pi_1/\pi_2$ for group probabilities $\pi_i$. This means that $\pi_1 = RR \pi_2$.

Thus, a power analysis to detect a relative risk of $RR$ is equivalent to a power analysis to detect a risk difference between $\pi_1$ and $RR \pi_2$. See chapter 3 in Biostatistical Methods 2e by John Lachin for more.

$\endgroup$
1
  • $\begingroup$ Thanks Demetri! $\endgroup$
    – RobertF
    Sep 14 at 14:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.