Estimating MA Parameters Through Factorization of the Autocovariance Matrix and an Ma‐Sieve Bootstrap
14 Pages Posted: 16 Apr 2018
Date Written: May 2018
A new method to estimate the moving‐average (MA) coefficients of a stationary time series is proposed. The new approach is based on the modified Cholesky factorization of a consistent estimator of the autocovariance matrix. Convergence rates are established, and the new estimates are used to implement an MA‐type sieve bootstrap. Finite‐sample simulations corroborate the good performance of the proposed methodology.
Keywords: ARMA models, sieve bootstrap, Wold representation
Suggested Citation: Suggested Citation