2
$\begingroup$

I am reading a paper and come across the following information:

 Predictor    Dependent Variable    R Square    Beta    P
     A                D                .12      .35   <0.05
     B                D                .16      .40   <0.05
     C                D                .13      .36   <0.05

Authors are using linear regression to compute the correlation between A and D, B and D and C and D, and they claim there is significant positive correlations between each and every one of them. I am confused and do not understand how can the authors draw such a conclusion with the presented data?

Improve this question
$\endgroup$
2
  • $\begingroup$ Which paper are you looking at? $\endgroup$
    – Glen_b
    Commented Nov 16, 2014 at 2:12
  • $\begingroup$ @Glen_b it's fossati.us/papers/fossati-its08.pdf in Section 4 (evaluation) fourth paragraph last sentence $\endgroup$
    – user1935724
    Commented Nov 16, 2014 at 2:54

1 Answer 1

Reset to default
3
$\begingroup$

The significance of the correlation between y and x is related to the significance of the coefficient in the regression of y on x.

Specifically, for the usual t-test on correlation under normality, they have the same p-value (they're said to be equivalent tests).

The conclusion that the correlation positive is based on the sign of the regression coefficient - in simple regression they have the same sign.

Improve this answer
$\endgroup$
7
  • $\begingroup$ You mean the p-value is less than 0.05 so it's a strong correlation? $\endgroup$
    – user1935724
    Commented Nov 16, 2014 at 2:53
  • $\begingroup$ No; statistical significance does not imply the correlation is strong, just that it's not consistent with a population correlation of zero. $\endgroup$
    – Glen_b
    Commented Nov 16, 2014 at 2:55
  • $\begingroup$ I guess I still not get what you mean. In this case how does the author conclude their claim? Because normally we look at R square, right? $\endgroup$
    – user1935724
    Commented Nov 16, 2014 at 2:57
  • 1
    $\begingroup$ My answer already explains how the correlations were concluded to be significantly different from 0 (that they're positive comes from the sign of the estimate of $\beta$). That covers the claim. That has nothing to do with concluding the correlation is strong, which was not mentioned in your statement of what they claimed. It's not clear to me where the difficulty lies. $\endgroup$
    – Glen_b
    Commented Nov 16, 2014 at 3:02
  • $\begingroup$ You mean any positive value of beta will make the correlation significant? $\endgroup$
    – user1935724
    Commented Nov 16, 2014 at 3:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.