I have estimated a multiple nonparametric regression using the loess command in R. I have also estimated an additive version of the model using the gam function. Is there a way of using the output of these two models to test the restrictions imposed by the additive model?