A method of estimating parameters in a model by minimizing the sum of squares of differences between observed and theoretical values of a variable. If
is a sample of
n observations, and μ
i is a set of theoretical values corresponding to a set of unknown parameters, θ, and a set of known associated observations,
xi, then the criterion to be minimized with respect to variations in θ is the sum of squares,
The values of θ at which the minimum occurs are known as
least squares estimates.
The method of weighted least squares is used when each observation is associated with a weight, wi (see measures of location), and the criterion to be minimized is
See also likelihood,
regression analysis.