Abstract
It is shown that if the probability integral of κ is the incomplete beta-function lκ(p, q) = α, the cube root of −logκ is distributed normally to good approximation for all q ≥ 2 and p ≥ q, at least for probabilities α between 0.05 and 0.95. This makes it easy to fit dose-response curves when α is the probability of response and κ depends on the dose, and to do so more simply than from another transformation of κ recently published in Biometrika.

This publication has 0 references indexed in Scilit: