Modelling Outliers and Missing Values in traffic Count Data Using the ARIMA Model.

This paper considers the application of the methodology to traffic count time series in which both missing values and outliers are present. Intervention analysis and detection using large residuals are shown to he reasonably effective but possible problems that result from non- stationarity in the data are identified. It is shown that despite considerable variabilty in the types of series the model selected from the ARIMA family is surprisingly homogeneous.