The Bayesian model evidence/ denominator of posterior sometimes could be way bigger than 1, which means the probability of observing a certain value under a hypothesis. How to interpret the meaning of evidence?
Asked
Active
Viewed 72 times
2
-
Obviously a value of $1$ is not a probability. That indicates you should take a closer look at exactly how your "model evidence/denominator of posterior" is computed and what it represents. I suspect the thread at https://stats.stackexchange.com/questions/4220 might be relevant. – whuber Jun 21 '18 at 12:07
-
If I understand correctly, do you mean that P(data|model(i)) is a continuous distribution with the variable model(i)? @whuber – TxWang Jun 21 '18 at 12:56
-
I don't mean anything specific because that aspect of your question is too vague for anyone but you to know what it refers to. I'm only guessing about the circumstances that might be motivating your reference to a value "way bigger than 1" as a "probability." – whuber Jun 21 '18 at 13:00