The Nadaraya-Watson estimator of regression is known to be highly sensitive to the presence of outliers in the sample. A possible way of robustication consists in using local L-estimates of regression. Whereas the local L-estimation is traditionally done using an empirical conditional...