math people

[gusl]

Isn't there a triangle-inequality result for correlation coefficients?

if CC(A,B)=.99 and CC(B,C)=.90 isn't there a lower bound on CC(A,C)? (.89?)

Likewise, if CC(A,B)=.99 and CC(B,C)=-.99, then there should be an upper bound on CC(A,C).


Is there a name for "1 - the correlation coefficient"?

Comments

[marknau.livejournal.com]

In your second case, A could be the same as C, so the upper bound is 1. I have heard the term "correlation distance" used for 1 - r, but have no idea how standard that is.

[marknau.livejournal.com]

Never mind, sorry, missed the minus sign on the second value.