Quick Correlation and Causality: A Scientist's Perspective

  • Context: Undergrad 
  • Thread starter Thread starter clane
  • Start date Start date
  • Tags Tags
    Correlation
Click For Summary
SUMMARY

The discussion centers on the relationship between correlation and causality, emphasizing that a correlation coefficient near zero does not definitively indicate the absence of causality. Participants clarify that while causality implies correlation, the reverse is not true; correlation can be linear or nonlinear. A zero correlation suggests that two variables are not linearly dependent, but they may still exhibit a non-linear relationship. Therefore, a zero correlation does not conclusively rule out causality unless non-linear relationships are also excluded.

PREREQUISITES
  • Understanding of correlation coefficients and their significance
  • Familiarity with linear and nonlinear relationships in statistics
  • Basic knowledge of elementary logic principles
  • Concepts of dependency between variables
NEXT STEPS
  • Research the implications of non-linear relationships in statistics
  • Study correlation vs. causation in depth using statistical tools like R or Python
  • Learn about advanced statistical methods for assessing causality, such as Granger causality tests
  • Explore the use of scatter plots and regression analysis to visualize relationships between variables
USEFUL FOR

Statisticians, data analysts, researchers, and anyone interested in understanding the nuances of correlation and causality in data analysis.

clane
Messages
1
Reaction score
0
Hello All,

I am trying to look at some correlation numbers and while I am aware that correlation does not always indicate causality can a correlation coeffiecent that is very near to zero be used to indicate that it is not a cause?

Thanks

C
 
Physics news on Phys.org
Elementary logic should give you the answer.

Causality => correlation.

Therefore: not correlation => not causality.
 
Let's be careful.

Causality does not always imply correlation because the relationship may be nonlinear. Correlation is a linear relationship.

If two variables are dependent, their relationship can be linear or nonlinear. If it's linear, their correlation will be significant. If their corr. is not significant, then you can say that the variables are not LINEARLY dependent on each other. But, they can still be non-linearly dependent.

So the answer is technically "No, a zero correlation does not mean no causality." A more roundabout way of putting it is: "A zero correlation would imply no causality only if a non-linear relationship can be excluded on other grounds."
 
Last edited:

Similar threads

  • · Replies 45 ·
2
Replies
45
Views
4K
  • · Replies 58 ·
2
Replies
58
Views
7K
  • · Replies 13 ·
Replies
13
Views
5K
  • · Replies 3 ·
Replies
3
Views
2K
  • · Replies 3 ·
Replies
3
Views
2K
  • · Replies 13 ·
Replies
13
Views
2K
  • · Replies 3 ·
Replies
3
Views
2K
  • · Replies 2 ·
Replies
2
Views
1K
  • · Replies 178 ·
6
Replies
178
Views
11K
  • · Replies 3 ·
Replies
3
Views
2K