On the Efficiency of the Hybrid Multivariate Least Trimmed Squares and Residual Autocovariance as a Robust Estimator for a Multivariate Autoregressive Model

Abstract

This study deals with autoregressive models and robust methods for multivariate time series data with serially correlated errors. Some of these methods are the multivariate least trimmed squares (MLTS) and the residual autocovariance (RA). In this study, a hybrid of the two former methods was developed. The researcher used an actual data with and without contamination for the comparison of methods and using R, a freeware. It was found that the developed hybrid method rather than an isolation of the multivariate least trimmed squares and residual autocovariance procedure, yields more efficient parameter estimates in terms of RMSE, relative efficiency, sum of squares error determinant, AIC and BIC. Keywords - Mathematical Statistics, Residual Autocovariance, Multivariate Least Trimmed Squares, Multivariate Time Series, Vector Autoregressive Model, Robust Method, Serially Correlated Errors, Philippines.