I am trying to present sample size determination based on determining a certain difference between two proportions and the power of the test. As I understand it, my calculation requires both z-alpha and z-beta. From past experience I am aware that z-beta for a power of 90% is 1.282 and z-beta for a power of 80% is 0.842.

However, I am not sure what z-beta is for a power of 70%. I would love to use this for my study as well and also would am curious as to if anyone can provide a table for these values or a way to find this in R. Thanks!


$z_\alpha$ and $z_\beta$ simply refer to the $\alpha$ and $\beta$ quantiles of the $N(0,1)$ distribution. What you are looking for is the qnorm() function:

[1] 1.281552

[1] 0.8416212

Therefore, quite simply:

[1] 0.5244005

Following this webpage's framework, an R function that calculates the required sample size given $p_A$ , $p_B$, $\alpha$, $\beta$ and $\kappa$ could be:

sample.size.diff.prop <- function(pA, pB, kappa, alpha, beta) {
nB <- (pA*(1-pA)/kappa+pB*(1-pB))*((qnorm(1-alpha/2)+qnorm(1-beta))/(pA-pB))^2
nA <- kappa*ceiling(nB)
cat("The required sample size is", ceiling(nA), "in group A and", ceiling(nB), "in group B.")}

Please note that the "beta" parameter is $1 - \textit{desired power}$. This little code snippet demonstrates the output generated by this function:

sample.size.diff.prop(pA = 0.65, pB = 0.85, kappa = 2, alpha = 0.05, beta = 0.2) 
The required sample size is 96 in group A and 48 in group B.
| cite | improve this answer | |

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.