This paper investigates the relative forecasting performance of some partially adaptive
estimators of ARIMA models based on flexible parametric error distributions which allow for
kurtosis and skewness. These methods are used to obtain forecasts of some well-known
data sets.