Is the significance of slope for linear regression the same thing as the significance of the correlation?
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
Yes, because the slope can be expressed as:
$$ \beta = \frac{cov(X, Y)}{var(X)} $$
While the correlation coefficient is:
$$ \rho= \frac{cov(X, Y)}{\sqrt{var(X)*var(Y)}} $$
Whatever significance test you do can be justified with an exact permutation test that shuffles the dependence of X on Y, which leaves $var(X)$ and $var(Y)$ as constants, so significance tests on both the slope and the correlation coefficient can be imagined as a significance test on the covariance of your two variables.
