Sign in

Least squares and the normal curve

1809Big moment

Legendre (1805) and Gauss (1809) publish the method of least squares, and Gauss ties it to the normal error curve. Fitting models to noisy data becomes routine.

The story

The method of least squares is a way to find the best-fitting line or curve through measurements that contain errors. It picks the values that make the sum of the squared differences between predictions and observations as small as possible. Adrien-Marie Legendre published the method in Paris in 1805 as an appendix to a work on comet orbits, and Carl Friedrich Gauss described it in 1809 in his Theory of the Motion of the Heavenly Bodies, claiming he had used it since 1795. The American Robert Adrain also published it in 1808, and the question of who deserves credit has been debated ever since.

Astronomers needed it. Their observations contained errors, and they had more measurements than unknowns, so no orbit fit them exactly. Gauss became famous in 1801 when he predicted where the dwarf planet Ceres, discovered by Giuseppe Piazzi and then lost in the Sun's glare, would reappear, and it was found near his predicted position. Legendre's method gave a simple rule for combining redundant data, and it built on probability theory and calculus, since minimizing the squared error means setting derivatives to zero and solving.

Gauss's major contribution in 1809 was to justify the method with probability. He showed that if errors follow what is now called the normal or Gaussian curve, the bell-shaped distribution, then least squares gives the most probable values, and the arithmetic mean is the best estimate for repeated measurements. Laplace's central limit theorem soon explained why the bell curve appears so often in practice. Within decades the approach was standard in astronomy and surveying, and later in economics, biology and social science through regression, a term from Francis Galton. Today the same principle powers statistics software and the linear models at the base of much of machine learning.

Why it mattered

  • Regression, the main tool of statistics, descends directly from least squares and the normal error curve.
  • Surveying, astronomy and navigation gained a systematic way to combine imperfect measurements into reliable results.
  • The pairing of least squares with the normal curve shaped the theory of statistical estimation for the next two centuries.
  • Modern data fitting, from econometrics to machine learning, still relies on the same squared-error idea.

Sources

  1. Least squares method Britannica
  2. Adrien-Marie Legendre (biography) MacTutor, University of St Andrews

Drafted by AI (Claude) from the sources above and checked against them by AI; no historian has reviewed this page. Spot a mistake? Suggest a correction. How pages are made

Community notes

Loading notes…

Learn more ranked by votes

Documentbritannica.com
Least squares method (Britannica)
Editor's pick
VideoSearch YouTube
least squares method history Gauss Legendre
Search YouTube
VideoSearch YouTube
least squares regression explained
Search YouTube