Statistical processing of traffic flow characteristics data

Abstract

In the course of statistical processing of traffic flows characteristics data, the check for the presence of anomalous measurements in the sampling should be done at the very start of processing. If anomalous measurements are detected, they should be excluded from the sampling at an early stage of the processing and not taken into considerations in further calculations. Numerous criteria have been developed to detect outliers, their effectiveness depends on the sample size. In practice, for technical and economic reasons, it is impractical to obtain a large number of measurements, as a rule the sampling should be processed on the basis of limited number of observations. In this regard, methods for detection of outliers with a small number of measurements, which include the method based on the use of the Romanovsky criterion, are of great importance. However, the analysis of literary references showed that in some recently published studies it is not recommended to use the Romanovsky criterion with the number of measurements less than 20. Therefore, the purpose of this study is to test the power of the Romanovsky criterion (test) for a small number of measurements and the possibility of its application in samplings of small size (n ≤ 20). The conducted studies have shown that the power of the Romanovsky criterion is quite high and it has high reliability with a small number of measurements, which makes it possible to use it in small samples to detect anomalous measurements.