Questions tagged [loess]

LOESS (or LOWESS) stands for locally weighted scatterplot smoothing.

It is a form of nonparametric regression (nonparametric in the sense of having an unspecified smooth form of underlying relationship between x and y, not the sense of being distribution-free). Specifically, it uses local (k-nearest neighbor) polynomial- (usually linear-) kernel regression. It also implements a robustness step where large outliers are downweighted.

Both terms were coined by Bill Cleveland, who developed them in the late 70s and into the 1980s; LOWESS is the older term, but LOWESS and LOESS are very similar and could be thought of as two implementations of the same concept.

Use this tag for questions relating to both LOESS and LOWESS.

Reference: Wikipedia Local regression