What’s So Funny About Correlation? xkcd.com
What is Correlation? When two things go together: cookies & milk macaroni & cheese peanut butter & jelly
What is Statistical Correlation? A mutual relationship between two variable quantities. Correlation coefficients measure how closely related the two variables are. Stronger relationships mean that a change in one variable is accompanied by a systematic change in the other. Correlation CoefficientStrength 1Perfect Positive 0.7 to 0.9Strong Positive 0.4 to 0.6Moderate Positive 0.1 to 0.3Weak Positive 0No Correlation -0.1 to -0.3Weak Negative -0.4 to -0.6Moderate Negative -0.7 to -0.9Strong Negative Perfect Negative
Does the amount of uranium stored at nuclear plants correlate with the amount of power they generate?
YearUraniumPower
Does the amount of uranium stored at nuclear plants correlate with the amount of power they generate?
Does the amount of uranium stored at nuclear plants correlate with the amount of power they generate?
Why might the amount of uranium stored at nuclear plants correlate with the amount of power they generate?
What is Causation? The action of causing something; the relationship between cause and effect. Smoking causes cancer Sugar causes cavities Gravity causes objects to fall
What is Causation? We can make an inference about a causal relationship based on the strength of correlation. We have to be careful about assuming an actual causal relationship, though.
How does consumption of chicken correlate with crude oil imports?
How does consumption of chicken correlate with crude oil imports? YearPer capita chicken consumption (lbs.) Total crude oil imports (millions of barrels) Chicken data from USDA. Oil data from Dept. of Energy.
How does consumption of chicken correlate with crude oil imports? 0.9
How does consumption of chicken correlate with crude oil imports?
Does eating more chicken cause crude oil imports to increase? ?
But wait… …if eating more chicken doesn’t cause crude oil imports to increase, then why are they so closely correlated? ?
The Lurking Variable Total US Population Year
Correlation and Causation Correlation is a measure of how well data fit a relationship. A correlated relationship between two quantitative variables is not the same as a causal relationship. Causal inferences about correlated events can be made based on underlying theories and knowledge. The determination that one thing causes another requires a scientific experiment.
"Correlation is not causation, but it sure is a hint.“ Edward Tufte, American Statistician
What’s So Funny About Correlation? Cartoon from xkcd.com