Nonparametric Monte Carlo Pricing of American Options

This paper presents a combined method based on non parametric regression and Monte Carlo algorithm to price America options. Simulated path are used to estimate the continuation value of the option. The value functions can then be computed by backward induction. Althought the flexibility of nonparametric regression allow to obtain accurate price estimates with remarkable speed some problems arise when we have to set the value of some technical parameters such as bandwidth, the order of polynomial, the number of simulated trajectories, the number of replications. The discussion are focused on these problems. Applications are provided on one and two dimensional american optons