In all likelihood, I shouldn’t be using probability at all, because probability theory doesn’t capture cause and effect well. Thinking back, what I should have said is just that rationalists are more likely to adopt polyamory than polyamorists are likely to adopt rationalism. The actual ratios of each are less relevant.
To be clear, this is almost the same as the formula you gave; I’m just using the log odds ratios formulation of Bayes theorem
LOR(X|E) = LOR(X) + log(P(E|X)) - log(P(E|NOT X))
where LOR(X) = log(P(X)/P(¬X))
in other words LOR(X|E) - LOR(X) = log(P(E|X)) - log(P(E|NOT X)) the log-likelihood ratio, the weight of evidence you need to update from one to the other.
I think P(X|E) - P(X) is the wrong measure—should be the log likelihood ratio log(P(E|X)) - log(P(E|NOT X))
I was feeling uncomfortable about that myself.
In all likelihood, I shouldn’t be using probability at all, because probability theory doesn’t capture cause and effect well. Thinking back, what I should have said is just that rationalists are more likely to adopt polyamory than polyamorists are likely to adopt rationalism. The actual ratios of each are less relevant.
To be clear, this is almost the same as the formula you gave; I’m just using the log odds ratios formulation of Bayes theorem
LOR(X|E) = LOR(X) + log(P(E|X)) - log(P(E|NOT X))
where LOR(X) = log(P(X)/P(¬X))
in other words LOR(X|E) - LOR(X) = log(P(E|X)) - log(P(E|NOT X)) the log-likelihood ratio, the weight of evidence you need to update from one to the other.