A method used for estimating parameters, particularly in regression analysis, by minimizing the difference between the observed response and the value predicted by the model.