Local & Nonparametric Regression

Abstract

This chapter is devoted to a class of regression procedures based on a new paradigm. Instead of searching for regression functions in a space whose elements are determined by a (small) finite number of parameters, we derive the values of the regression function from local properties of the data. As we shall see, the resulting functions are given by computational algorithms instead of formulae in closed forms. As before, we emphasize practical implementations over theoretical issues, and we demonstrate the properties of these regression techniques on financial applications: we revisit the construction of yield curves, and we propose an alternative to the Black-Scholes formula for the pricing of liquid options.