Forecasting in least absolute value regression with autocorrelated errors: a small-sample study

Abstract

Least absolute value (LAV) regression is a robust alternative to ordinary least squares (OLS) and is particularly useful when model disturbances follow distributions that are nonnormal and subject to outliers. The performance of the OLS estimator when the disturbances are autocorrelated has been studied extensively, but the performance of the LAV estimator in the presence of serial correlation is less well established. In this research, we study the forecasting performances of OLS- and LAV-based models for simple time series regression when the errors are autocorrelated. Monte Carlo simulation methods are used to compare the forecasting accuracies of the different models. A least absolute value analogue of the Prais-Winsten correction possesses an appealing robustness for the context under consideration.