1
$\begingroup$

I have 10 independent variables (IV) that may predict my dependent variable. There's a lot of multicollinearity in my data (r between IVs is r=0.4 on average but not higher than r=0.8). I suspect that's because of layered effects: Like IV2 and IV3 directly influence the dependent variable, but IV2 itself is influenced by IV4 and IV7.

I'm looking for the right terminology for search: Which keywords/methods can help me to interpret the causality of my model and the layered effects structure?

Improve this question
$\endgroup$
3
  • $\begingroup$ What is your causal model? $\endgroup$
    – Alexis
    Commented Dec 27, 2022 at 17:31
  • $\begingroup$ Do you have a causal model/graph or are you asking how to/whether you can build one from your data? $\endgroup$
    – dipetkov
    Commented Dec 27, 2022 at 17:46
  • $\begingroup$ I am asking how to / whether I can build a causal model from my data. (I also have the literature that suggest a certain causal model.) $\endgroup$
    – CrossVal_Asker
    Commented Dec 27, 2022 at 20:30

1 Answer 1

Reset to default
0
$\begingroup$

If one or more independent variables are an exact linear combination of other independent variables, then the model suffers from perfect collinearity, and it cannot be estimated by OLS. This assumption does allow the independent variables to be correlated, they just cannot be perfectly correlated. The solution to perfect collinearity is simple: drop any one of the variables for which you suspect perfect collinearity. High but not perfect correlation between two or more independent variables is called multicollinearity: multicollinearity violates none of OLS assumptions, a rule of thumb is usually to exclude covariate with a linear correlation coefficient higher than .8. You may consider to calculate the Variance Inflation Factor to have a deeper insight.

Improve this answer
$\endgroup$
4
  • $\begingroup$ thanks for your answer! Well they are not prefectly correlated, the highest linear correlation coefficient is 0.7. $\endgroup$
    – CrossVal_Asker
    Commented Dec 27, 2022 at 17:25
  • $\begingroup$ Your distinction between collinearity and multicollinearity is false. This has to do with how many variables are involved, not whether it is perfect or merely approximate. $\endgroup$
    – Richard Hardy
    Commented Dec 27, 2022 at 20:14
  • $\begingroup$ @RichardHardy please check your information. See Introductory Econometrics: A Modern Approach, Textbook by Jeffrey Wooldridge chapter 3 page 80, 81, 82, 90 for the distinction between collinearity and multicollinearity. $\endgroup$
    – Barbab
    Commented Dec 28, 2022 at 0:26
  • $\begingroup$ To agree with @RichardHardy, the numerical distinction (i.e., how many variables) is fairly standard in statistics. In economics, however, a modifier helps clarify the two concepts, but for reasons more closely aligned with your perspective. For example, the distinction between perfect collinearity and multicollinearity is, in my view, a little less subtle. $\endgroup$
    – Thomas Bilach
    Commented Dec 29, 2022 at 4:05

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.