He only talked about correlation, not causation. The most likely causation is indeed the one you posited.
EDIT: ignore the following.
But two things that are both (positively) correlated with a third are (positively) correlated with each other, no matter the the direction or even existence of causal relations.
But two things that are both (positively) correlated with a third are (positively) correlated with each other, no matter the the direction or even existence of causal relations.
I don’t believe this is the case. Two things things both being positively correlated with a third are more likely to be correlated with each other, all things being equal. Yet there are causal relations which could make those things negatively correlated with each other while both positively correlated with the third. The most obvious examples would be of partisan behaviors where the ‘third’ is a generic factor that encourages someone to pick a side.
I don’t think “are generally” applies “no matter the the direction or even existence of causal relations”.
If A causes E and B independently causes E, then there will be correlation between A and E and between B and E, but no reason to expect correlation between A and B.
You’re right, and I really should have known better. This is one of the examples used in Judea Pearl’s Casaulity, about how to assign plausible causation structures given only correlations.
He only talked about correlation, not causation. The most likely causation is indeed the one you posited.
EDIT: ignore the following.
But two things that are both (positively) correlated with a third are (positively) correlated with each other, no matter the the direction or even existence of causal relations.
I don’t believe this is the case. Two things things both being positively correlated with a third are more likely to be correlated with each other, all things being equal. Yet there are causal relations which could make those things negatively correlated with each other while both positively correlated with the third. The most obvious examples would be of partisan behaviors where the ‘third’ is a generic factor that encourages someone to pick a side.
You’re right. I should have said “are generally”, rather “are”.
I don’t think “are generally” applies “no matter the the direction or even existence of causal relations”.
If A causes E and B independently causes E, then there will be correlation between A and E and between B and E, but no reason to expect correlation between A and B.
You’re right, and I really should have known better. This is one of the examples used in Judea Pearl’s Casaulity, about how to assign plausible causation structures given only correlations.
A pair of correlations between A and B, and between B and C, is correlated with a correlation between A and C. :P