Nonlinear Filtering for Non-Stationary Multivariate Cointegration Models Author:

The study of non-linear filtering for non-stationary multivariate cointegration models combines the disciplines of engineering and time series analysis. Within this thesis we consider different methods of filtering for multivariate cointegration models and also techniques to estimate the parameters of the models. The effectiveness of the filtering and parameter estimation techniques are investigated within a series of case studies utilising novel implementations of the filtering and estimation algorithms in Object-oriented MATLAB. Finally, multivariate cointegration model estimation techniques are applied to a set of empirical data, consisting of futures contract pairs of indices, interest rates and bonds