Inference for Stationary Processes Using Banded Covariance Matrices
We consider prediction and estimation problems by banding covariance matrices of stationary processes. Under a novel short-range dependence condition for a class of nonlinear processes, it is shown that the banded covariance matrix estimates converge in operator norm to the true covariance matrix with reasonable rates of convergence. A sub-sampling approach is proposed to choose the banding parameter.
KeywordsTime Series Data Covariance Matrice Toeplitz Matrice Nonlinear Time Series Covariance Matrix Estimate
Unable to display preview. Download preview PDF.
- Bickel, P. J. and Levina, E.: Regularized estimation of large covariance matrices. (2006). www.stat.berkeley.edu/∼bickel/techrep.pdf
- Tong, H.: Non-linear Time Series: A Dynamical System Approach. Oxford Scienti c Publications. (1990).Google Scholar
- Zeitouni, O. and Anderson, G.W.: A CLT for regularized sample covariance matrices. Ann. of Statist. To appear. (2008).Google Scholar