Correlation


Correlation is the statistical association between two (or more) equal variables.

  • does not tell us that one influences the other
    • but it does indicate that if one variable changes, the other variable changes as well
  • simply attempting to determine what relationship might exist between two variables
    • not attempting to prove that one variable caused the other

Causal relationship is a relationship in which one variable is proven to have an effect on another.

  • in order to investigate whether a causal relationship exists, a correlation coefficient must be calculated

Pearson’s correlation coefficient is a calculation used to measure a linear relationship between data points, returning a value that is plus or minus one to determine the strength of the relationship.

  • measures the linear relationship between two numerical (continuous) variables
  • is expressed as an r value
    • A value that when closer to one shows a strong correlation between values and when closer to 0 means no correlation
    • R values between 0.4 and 0.7 represent a moderate correlation
  • coefficient of determination is expressed as R2
    • square of the correlation coefficient value
    • used to interpret the determination
    • easier to understand if you multiply it by 100%

Example

Let’s see this in action by using the student scores and absenteeism example from our simple linear regression output.

Regression Statistics
Multiple R0.48512063
R Square0.23534203
Adjusted R Square0.20475571
Standard Error11.0981969
Observations27
  • The Multiple R is Pearson’s Correlation coefficient (r) and is currently showing 0.48
    • means that there is a moderate correlation between the test score and the days absent
  • looking at R Square value, we can theorize that the 23% of variability in the score is explained by the variability in the days absent
    • means that 77% of the variability is explained by other data not expressed in this sample
  • tells us that even though there appears to be a relationship between the number of days absent and the test score,
    • the strength of that relationship is not high enough to explain or consider absences as the only factor affecting test scores