a form of
regression analysis in which a model of the relationship between outcomes and predictors is obtained by fitting different linear or quadratic functions to different segments or intervals of data. Variables are assigned different
weights in the
regression equations to reflect their relative importance at each point and no assumptions are made about the associations among the variables. The overall curve obtained by combining the individual fitted curves for the different data segments shows the general shape of the relationship between the variables. Also called
locally weighted regression.