Patterns in Data: Correlation, Trends & Outliers
Find the signal hiding in the noise
Find the signal hiding in the noise
Each dot is a data point with two values (x and y). Click to add points. A trend line automatically fits to show the relationship. Is it positive, negative, or none?
Click to add data points, or load a preset pattern.
These scatter plots show real correlations that do NOT mean one thing causes the other. Click each to see why.
The amount of cheese consumed per person in the US correlates 95% with the number of civil engineering doctorates awarded each year. Obviously cheese does not cause PhD degrees. Silly correlations like this are everywhere in data if you look hard enough.
Click the red outlier point to add it to the dataset. Watch how a single extreme value changes the entire trend line.
Without the outlier: strong positive trend. Add it and watch the line shift.
The skills you just practiced apply to any field that uses data.
Does this drug reduce symptoms? Researchers plot treatment vs outcome and look for correlation. Randomized trials establish causation.
GDP vs unemployment, inflation vs interest rates. Economists find patterns in economic data to guide policy decisions.
Moneyball: the Oakland A's found that on-base percentage correlated with wins better than batting average. Data patterns won games.
You've learned to see patterns, trends, and outliers in scatter plots. You know that correlation does not imply causation, and that a good trend line reveals the story data is trying to tell.
When two variables move together (both go up, or one goes up while the other goes down), they are correlated. The closer to a line, the stronger.
Ice cream sales and drownings both rise in summer. Ice cream does not cause drowning. They share a cause: hot weather. Always ask WHY before assuming cause.
One extreme point can dramatically change a trend line. Always check if your pattern survives removing the most extreme values.
A line of best fit extends the pattern into the future. But extrapolating too far is dangerous. Past patterns do not guarantee future results.
Put your new knowledge into practice!