There are quite a few variants of the Cauchy-Schwarz inequality and I'm saying this because there is one related to statistics as well and it goes like this

Answer by Daeyoung Lim:

There are quite a few variants of theCauchy-Schwarz inequalityand I'm saying this because there is one related to statistics as well and it goes like this:[math]\operatorname{Cov}^{2}\left(X,Y\right) \le \operatorname{Var}\left(X\right)\operatorname{Var}\left(Y\right)[/math]It follows from this Cauchy-Schwarz inequality that the correlation coefficient is between -1 and 1.[math]-\sqrt{\operatorname{Var}\left(X\right)\operatorname{Var}\left(Y\right)} \le \operatorname{Cov}\left(X,Y\right) \le \sqrt{\operatorname{Var}\left(X\right)\operatorname{Var}\left(Y\right)}[/math]Therefore,[math]-1 \le \frac{\operatorname{Cov}\left(X,Y\right)}{\sqrt{\operatorname{Var}\left(X\right)\operatorname{Var}\left(Y\right)}}\le1[/math]The term in the middle is exactly the correlation coefficient!Actually the covariance is the inner product between two random variables and the standard deviation is the norm of a random variable. If we denote the inner product [math]<X,Y>[/math] and the norm [math]|X|[/math], then the usual Cauchy-Schwarz inequality still holds:[math]<X,Y>^{2}\le |X|^{2}|Y|^{2}.[/math]But then the inner product is also defined as follows:[math]<X,Y>=|X||Y|\cos\left(\theta\right)[/math]The correlation coefficient is in fact the cosine of the angle between two variables:[math]\operatorname{Corr}\left(X,Y\right) =\frac{<X,Y>}{|X||Y|}=\cos\left(\theta\right)[/math]It's obvious that cosine ranges only from -1 to 1, right?

Why is the correlation coefficient between -1 and 1?

Advertisements