Local Regression: The LOESS Procedure

PROC LOESS implements a local regression approach for estimating regression surfaces that was pioneered by Cleveland, Devlin, and Grosse (1988). No assumptions about the parametric form of the entire regression surface are made by PROC LOESS. Only a parametric form of the local approximation is specified by the user. Furthermore, PROC LOESS is suitable when there are outliers in the data and a robust fitting method is necessary.