CARMA Processes driven by Non-Gaussian Noise
Abstract
We present an outline of the theory of certain L\'evy-driven, multivariate stochastic processes, where the processes are represented by rational transfer functions (Continuous-time AutoRegressive Moving Average or CARMA models) and their applications in non-Gaussian time series modelling. We discuss in detail their definition, their spectral representation, the equivalence to linear state space models and further properties like the second order structure and the tail behaviour under a heavy-tailed input. Furthermore, we study the estimation of the parameters using quasi-maximum likelihood estimates for the auto-regressive and moving average parameters, as well as how to estimate the driving L\'evy process.
Cite
@article{arxiv.1201.0155,
title = {CARMA Processes driven by Non-Gaussian Noise},
author = {Robert Stelzer},
journal= {arXiv preprint arXiv:1201.0155},
year = {2012}
}
Comments
Preprint version of article available at http://www.tum-ias.de/institute-for-advanced-study/publications/journal-primary-sources.html