There Are No Magic Outcome Variables

Thanks for sharing, read the blog post but not the underlying paper, a quick side thought - ratio outcomes are seriously under-taught compared to proportions (which are a special case in some sense. This might come from the “coin flips all the way down” learning path).

But I don’t understand why either of them did ln(G)/P instead of ln(G/P), wtf? I don’t have much intuition for what that is supposed to measure. Some algebra:

Y = ln(G)/P
Y = ln(\sqrt[P]{G})
\exp{Y} = \sqrt[P]{G}

So it would be like doing geometric means instead of arithmetic, except that G is an sum and not a product. So what is it?