Regression

First taught in High school statistics

Definition

Regression is a method for finding an equation that best describes how one variable depends on another, so you can make predictions. Linear regression fits a straight line.

02468100246810xy
The regression line makes the dashed vertical distances (residuals) as small as possible overall.
Advertisement

The most common regression line is the least-squares line

where ("y-hat") is the predicted value, is the slope and is the Y-intercept. It is chosen to make the sum of the squared residuals as small as possible.

Example: A regression of test score on hours studied gives .

  • The slope means each extra hour of study predicts 4.5 more points.
  • For 6 hours, the predicted score is .

Use regression to predict only within the range of the data. Predicting far outside it (extrapolation) is risky. The correlation coefficient tells how well the line fits.

Advertisement