LOESS

LOESS(x;d;s;l) or LO(x;d;s;l) fits a locally weighted regression of order l (= 1 for linear, 2 for quadratic) with approximately d degrees of freedom or using smoothing parameter s (regression models only): x is a variate for univariate smoothing, or a pointer to up to four variates for multivariate smoothing; when x is a variate l is a scalar, when x is a pointer it is either a scalar or a variate with an element for each variate in the pointer (regression and generalized linear models only).