Key points I have in mind, and then followed by my question: 1. I have a regression linear model with a set of attributes and their coefficients. 2. I also ran a correlation analysis on these attributes to check for any correlations 3. question: Assume I am inspecting the correlation between 2 variables (considered independent variables in this case, when we look at correlation): if corr = -0.22, how would the intercept factor in? or would it at all? Y~a1X+a2V+ b Does this correlation coefficient include any indication about the intercept?

  • $\begingroup$ At stats.stackexchange.com/a/108862/919 I have explained how the full covariance matrix of all three variables determines the regression coefficients, but not the intercept. $\endgroup$ – whuber Sep 26 '18 at 16:28

The correlation won't tell you anything about the intercept. The intercept can be interpreted as the average value of Y when X = 0. So, by definition the sign of the correlation between X and Y does not have any influence on the intercept of Y, since X is zeroed out.

  • 1
    $\begingroup$ You needn't be so tentative: the correlation won't tell you anything about the intercept, period. That's because shifting the response by adding any constant to it will not change the correlation matrix but it will add that constant to the intercept. $\endgroup$ – whuber Sep 26 '18 at 16:33
  • $\begingroup$ whuber, thanks for your comment. I edited my response accordingly. $\endgroup$ – Sympa Sep 27 '18 at 1:03

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.