SAODR: sequence analysis for outlier data rejection

In automatic data acquisition, a sample is generally made up of several instrumental readings. A series of readings is generally reduced to a single value by simple methods, such as averaging. However, outlying values can affect the series. The paper introduces an algorithm, named 'sequence-analysis outlier data rejection' (SAODR), which takes into account one of the most common problems affecting the measurand during the acquisition, i.e. a nonlinear drift with embedded sequences of outliers due to pulse-noise peaks. The algorithm uses a time-ordering procedure and the 'distances' between successive readings. The frequent case of constant sampling rate is discussed. The reported tests show the results obtained with Fortran 77 and MATLAB® implementations of the algorithm. A rejection efficiency higher than 99% was obtained.