Comparative Study of Estimation Methods of The Parameters of VMA(1) Via Simulation and Construction of Monte-Carlo Confidence Intervals

Four methods of estimation of parameters of two components vector moving average model
VMA(1) are compared. These methods are exact maximum likelihood (via Kalman filtering), Yule Walker type, moments estimation and Godolphin type. All estimations are based on simulated data using two different covariance matrices. All methods are asymptotically equivalent well inside the invertibility region. Method of constructing Monte-Carlo confidence intervals for parameters is suggested for the time series of the fitted model.