Bayesian Non-Linear Modellings of the Short Term US Interest Rate: the Help of Non-Parametric Tools.

This paper is concerned with the empirical investigation of models of the US short term interest rate, using a mixture of classical non-parametric methods and of Bayesian parametric methods. The shape of the drift and volatility functions of the usual di usion equation are rst investigated using a preliminary non-parametric analysis. The paper then develops a Bayesian method for comparing models which is based on the ability of a model to minimise the Hellinger distance between the posterior predictive density and the density of the observed sample.