Detecting outlying observations and structural changes in European air quality data
ETC/ACM Technical Paper 2012/16

Released: 2013/05/22: See the report

This paper describes approaches for the detection of outlying observations and structural changes in European air quality data. Typical artefacts that can occur are outlying observations and break points and identifying such 'inhomogeneties' is useful both for causal research and quality improvement. Recommended moving window statisticse are used for outlier detection and Kolmogorov-Zurbenko fi lters for break detection. Both are of explorative nature and do not provide any probability statements about identied (i.e. potential) inhomogeneities.
For outlier detection is suggested to use the Tukey heuristic as it does not assume the data to follow a particular distribution. For break detection it is suggested that the method is rather robust against the choice of iteration parameters, while using narrow window widths clearly increases precision of locating the break.

Reference to R-scripts, data and other materials used for the calculations:

Prepared by: ETC/ACM subcontracted Institute for Geoinformatics (IfGI), University of Münster, Germany: Mirjam Rehr, Edzer Pebesma, Benedikt Gräler.

Published by: ETC/ACM, May 2013, 49 pp.