0
$\begingroup$

I have a situation where IVs positively correlate with the DV, and the IVs also correlate with eachother.

I do not explain how the beta weights are negative. No multicollinearity is detected, and the model is justified, and has an r-squared value of 5. Note for example, issue with Lev and Size; both correlate do DV CSRDI, both correlate to eachother, but negative beta for LEV and low significance?

This question specifically is asking about sign reversal. I am confused with this due to the lack of a multicollinarity issue. Lev and Size are correlated but not excessively, so why do IVs such as lev have a sign reversal in the regression model? Also is this classified as suppression? I am in no way a statistics expert and am only interested in understanding this for an assignment for my undergrad degree. So if possible, could you please explain in laymans terms. Thanks you

I was wondering if you anyone can explain this, and perhaps help with interpretation Alex Williams Correlations Alex Williams - Regression

Improve this question
$\endgroup$
5
  • $\begingroup$ You should find the answer in the linked thread. Please read it. If you still have a question afterwards, come back here & edit your Q to state what you have learned & what you still need to know. Then we can provide the information you need without just duplicating material elsewhere that already didn't help you. $\endgroup$
    – gung - Reinstate Monica
    Commented Feb 23, 2015 at 21:47
  • $\begingroup$ The correlation between Lev & Size is .438. That may not be 'problematic' multicollinearity, but it could still cause the sign to flip in the same way as described at the linked thread. $\endgroup$
    – gung - Reinstate Monica
    Commented Feb 23, 2015 at 22:51
  • $\begingroup$ @whuber, do you see something in this thread that I don't? This is a FAQ. I don't see anything here to be explained that doesn't already exist elsewhere on the site. The linked thread does not even mention multicollinearity. $\endgroup$
    – gung - Reinstate Monica
    Commented Feb 23, 2015 at 22:56
  • $\begingroup$ @gung I do not see there any explicit answer to this question, which asks us to compare correlation coefficients to multiple regression coefficients. Although I am in complete agreement that answers to this question have appeared (in many places), I am not sure that it is sufficiently obvious to a neophyte that those threads truly address the same question. You could convince me otherwise by providing a link to an obvious duplicate. $\endgroup$
    – whuber ♦
    Commented Feb 23, 2015 at 23:01
  • 1
    $\begingroup$ @whuber, do you think Positive correlation and negative regressor coefficient sign, Positive correlation but negative coefficient in regression, Is the model wrong if a coefficient changes from minus in correlation table to plus in OLS?, or Positive correlations to dependent variable, but negative coefficients are clear enough? I can add another answer here, if need be. $\endgroup$
    – gung - Reinstate Monica
    Commented Feb 23, 2015 at 23:23

0

Reset to default