Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Is there a generic answer as to how the log transform affects the Pearson correlation between two variables, i.e., is $\rho(\log(x),\log(y)) < \rho(x,y)$ always or the other way around and why?

share|improve this question
I don't have proof, but it can be both. – Julius Sep 9 '12 at 0:36
I think there may be a way to generalize, something to do with the skew of x relative to y. I obviously am not sure which is why I posted. – ganesha Sep 9 '12 at 0:44
My general tack when I ask myself this kind of question, Is to do some rapid numerical simulation in R. Make a habit of it, it will teach you a lot! – kjetil b halvorsen Sep 10 '12 at 19:49

Perhaps something can be said using other statistics, but there isn't a universal relationship of the form you specify in the question.

Suppose the joint distribution of $(x,y)$ is uniformly distributed on $\lbrace(0.1,1),(1,0.1),(.0001,.0001) \rbrace$. $\rho(x,y) = -0.335$, but $\rho(\log(x),\log(y)) = +0.885$.

Suppose $(x,y)$ is uniformly distributed on $\lbrace(.0001,1),(1,.0001),(10,10) \rbrace.$ Then $\rho(x,y) = +.984$, but $\rho(\log(x),\log(y)) = -.143$.

Variations on these examples can make one correlation close to $+1$ while the other is close to $-1$. The difference between $0.1$ vs. $0.0001$ might not make a large difference when you calculate $\rho(x,y)$, but it changes $\log x$ and perhaps $\rho(\log(x),\log(y))$ by a lot.

share|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.