Correlation

Creator
Creator
Seonglae Cho
Created
Created
2021 May 31 5:44
Editor
Edited
Edited
2024 Nov 30 11:39

Normalized
Covariance
, measures linear correlation

The correlation coefficient, Pearson product-moment correlation coefficient, PPMCC, Bivariate correlation, Pearson's r
Correlation does not imply causation such as spurious correlation.
Since variance is non-negative, correlation bounds to -1 to 1 (V(XσX±YσY)=2(1±ρ)V(\frac{X}{\sigma_X} \pm \frac{Y}{\sigma_Y}) = 2(1\pm\rho))
1ρXY=corr(X,Y)=cov(X,Y)σXσY1-1 \leq \rho_{XY} = corr(X, Y) = \frac{cov(X, Y)}{\sigma_X\sigma_Y} \leq 1corr(X,Y)=1Y=aX+bcorr(X,Y) = 1 \leftrightarrow Y = aX + b
Uncorrelated corr(X,Y)=0corr(X, Y) = 0 This means that there is no linear relationship.

Difference with
Convolution
is flip

(fg)(x,y)=f(i,j)I(x+i,y+j)(f \ast g) (x, y) = \int f(i,j) I(x+i, y+j)
Uncorrelated does not imply independence. However, if two random variables X and Y are
Normal distribution
uncorrelated, then they are
Pairwise Independence
. Since zero correlation makes
Covariance Matrix
and thus inverse of covariance matrix become diagonal. The joint PDF simplifies to the product of two independent normal distributions which satisfies
Pairwise Independence
.
The correlation between an
Even function
and the original variable is zero. (Integral cancels out) Naturally, even powers also exhibit this property.
Correlations
 
notion image
 
 
 
 
 

Recommendations