In statistics, linear regression is a technique for estimating the relationship between an independent variable, X, and its scalar result, the dependent variable, Y, derived from a series of X-Y relationships.

The computational routine involves trying to fit a straight line between a scatter plot of X-Y coordinates such that the sum of the squared differences between each dependent outcome, Yi, and the line computed at each Xi is minimized. This is the so-called "ordinary least squares (OLS)" estimator.

R-squared, often called the "coefficient of determination," expresses the "goodness of fit" of the line to the scatter plot. It ranges from 0, no fit whatever, to 1, which is a perfect relationship where all the points in the plot fit along the straight line.