The np package contains a function called npconmode which performs 'kernel modal regression'. Does anyone know of a reference that explains this? The R documentation cites Li and Racine's book, but I couldn't find any reference to kernel modal regression in it. I gather that this is the preferred function for a nonparametric regression with a discrete dependent variable. Based on Li and Racine, I would have thought that a conditional conditional density estimation (i.e. npcdens) was the way to accomplish this. I'd like to understand the difference. thanks, b. wagman