I would be interested to see whether computing P(A∩B) falsely to the average of P(A) and P(B|A) would model the error well. Like this any detail B that fits well to the very unlikely primary event A increases its perceived likelihood.
I would be interested to see whether computing P(A∩B) falsely to the average of P(A) and P(B|A) would model the error well. Like this any detail B that fits well to the very unlikely primary event A increases its perceived likelihood.