4
$\begingroup$

The Pearson correlation coefficient is the geometrical mean of the slopes of the two regression lines. For me it is difficult to understand how a geometrical mean of the slope of the two regression lines can be a measure of the linear correlation between two variables X and Y.

I found this paper fantastic, Bravais-Pearson and Spearman correlation coefficients: meaning, test of hypothesis and confidence interval by Artusi and coll., in explaining the Pearson coefficient, but I have difficulties in doing the last deduction.

$\endgroup$
2
$\begingroup$

Let's regress $Y$ on $X$ and write $b_{0x} + b_{1x} x$ for the regression line, where $b_{1x}$ is the slope. Then Pearson correlation coefficient is $r = b_{1x} \frac{s_x}{s_y}$, $s_x$ and $s_y$ being the standard deviations of $X$ and $Y$, respectively. If, on the contrary, we regress $X$ on $Y$, the regression line is defined by $b_{0y} + b_{1y} y$, thus $r = b_{1y} \frac{s_y}{s_x}$.

Then it follows that $r^2 = b_{1x} b_{1y}$, i.e. $r = \sqrt{b_{1x} b_{1y}}$, which is the geometric mean of the two slopes.

$\endgroup$
1
$\begingroup$

This is an interesting article with different interpretation of the corr. coeff. https://www.stat.berkeley.edu/~rabbee/correlation.pdf This does not give an answer to your question, but maybe you can find other interpretation

$\endgroup$
1
  • 1
    $\begingroup$ Thank you for your answer. But also there I cannot find the last deduction that I have asked $\endgroup$ – Ashish Bhigah Dec 6 '20 at 19:12

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.