Continuous-time Identification of Exponential-Affine Term Structure Models

This thesis addresses the problem of parameter estimation of the exponentialaffine class of models, which is a class of multi-factor models for the short rate. We propose a continuous-time maximum likelihood estimation method to estimate the parameters of a short rate model, given set of observations that are linear with respect to the interest rate factors. We assume that observations are corrupted by Gaussian noise with a known covariance, which lead to a maximum likelihood estimation method for partially observed systems. Unlike other approaches in the literature, we do not discretize either the interest rate model or the observation model.