This book presents a series of dynamic models of the term structure of interest rates, covering both theory and estimation in a unified framework. Special emphasis is placed on models which are driven by innovations that have a Gaussian mixture distribution. These models are able to flexibly capture the observed non-normality in the distribution of bond yields. It is shown that the theoretical models can easily be cast into the statistical state space form, which provides a convenient framework for statistical inference. An application to US data illustrates the properties of the models and shows the estimation techniques at work.