0
$\begingroup$

I was wondering whether correlation between $2$ variables inform us about their standard deviation.

Let's say $X$ has a standard deviation of $\sigma_1$ and $Y$ has a standard deviation of $\sigma_2$.

Group 1 - Made of equal elements of $X$ and $Y$ being negatively correlated with each other.

Group 2 - Made of equal elements of $X$ and $Y$ positively correlated with each other.

Is is possible to claim which group has a lower standard deviation or variance?

$\endgroup$
  • $\begingroup$ What is the variance of a group ?? $\endgroup$ – Yves Daoust Mar 16 '18 at 8:09
  • $\begingroup$ Notice that $(X,Y)$ and $(X,-Y)$ have opposite correlations. $\endgroup$ – Yves Daoust Mar 16 '18 at 8:15
1
$\begingroup$

By definition

$$\text{Corr}(X,Y)=\frac{\text{Cov}(X,Y)}{\sigma_X\sigma_Y}$$

We cannot conclude anything about the standard deviations from the correlation since correlation depends on covariance $and$ standard deviation.

In fact, the standard deviations are only used to scale the covariance so that the correlation will lie between $-1$ and $+1$, improving interprebility.

Here is an example in R statistical software:

Suppose we multiply $X$ by a scalar.

x <- rnorm(10^6, mean=5, sd=3)
y <- rexp(10^6, rate=1)
cor(x, y)
cor(5*x, y)  
sd(x)
sd(5*x)

This gives

$$\text{Corr}(X,Y)=7.83\cdot10^{-5}$$

$$\text{Corr}(10\cdot X,Y)=7.83\cdot10^{-5}$$

but

$$\sigma_X\approx3$$

$$\sigma_{10X}\approx15$$

so increasing the standard deviation did not change the correlation!

$\endgroup$
  • $\begingroup$ Depends on independence between the 2 events too. $\endgroup$ – Aravind SS Mar 17 '18 at 16:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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