I wonder, of the many distinctions made between these statistical notions whether the following interesting one has been made, i.e., to declare two random variables independent, an observer is not required, whereas only an observer can pronounce two random variables as correlated, post-hoc. Hence, independence is a property of the distribution, (in the context of generative function as opposed to histogram), whereas correlation is a property of one single realization sampled from that distribution with respect to another such realization.
If I were Hugh Macleod, I would say at this point, "Exactly. Bayesian vs. Frequentist."
1 year ago