Variance Estimation in Time Series Regression Models

摘要: The effect of variance estimation regression coefficients when disturbances are serially correlated in time series models is studied. Variance enters into confidence interval estimation, hypotheses testing, spectrum and expressions for the estimated standard error prediction. Using computer simulations, robustness various estimators, including Estimated Generalized Least Squares (EGLS) was considered. estimates coefficient estimators produced by packages were Models generated with a second order auto-correlated structure, considering based upon misspecified order. Ordinary (OLS) (order zero) outperformed first EGLS. A full comparison zero four indicate that over specification preferable to under specification.