Contributions of influence function using the inverse autocorrelation function in the detection of outliers

Abstract

Outliers in time series, depending on their nature may have a moderate to significant impact on the effectiveness of the standard methodology for time series analysis with respect to model identification, estimation and forecasting. The suggested procedure used for identifying the outliers graphically in time series data was investigated by considering the influence function for the inverse autocorrelation function (IACF). Form the findings, it was noticed that for large series the influence was almost positive in values while for relatively short series the large negative influence are noticeable. The model order determination technique was also proposed.

JONAMP Vol. 11 2007: pp. 627-634