Alternative Methods of Dealing with Outliers in Forecasting Sales with Regression-Based Models

Sales time series are frequently subjected to distortions caused by advertising, competition, promotions, accounting systems, and stockouts. These distortions are called outliers in statistical literature. In recent years, there has been an increasing number of articles dealing with the forecasting of sales using better time series models. In addition to building better models, a second way to improve forecasting accuracy is to improve the source data by a careful analysis of the process which generated the data.