The point of the "method of least-squares" is to create a line that
best fits a given set of nonlinear points. In most people's
experience, this is done by eye, but there is a more mathematical
way to do it. Each point has a vertical distance away from the line
that's supposed to fit it. The line created by the least squares
method seeks to make the totals of those distances as small as
possible. When a particular combination of y-intercept (b) and
slope (a) create the lowest total possible, the least squares
regression line is created, with the equation y=ax+b. Explore this
concept, every red point can be moved.
sketch created by Bill Finzer, March 1995, September 1997