Detecting outliers in irregularly distributed spatial data sets by locally adaptive and robust statistical analysis and GIS |
| |
| Authors: | Hongxing Liu Kenneth C Jezek Morton E O'Kelly |
| |
| Institution: | 1. Department of Geomatics , The University of Melbourne , Victoria 3010, Australia E-mail: winter@unimelb.edu.au;2. Department of Spatial Information Science and Engineering , University of Maine , 5711 Boardman Hall, Orono, ME 04469-5711, USA E-mail: nittel@spatial.maine.edu |
| |
| Abstract: | In this paper, we propose a new method for detecting outliers in an irregularly distributed spatial data set. Our method has two desirable properties. First, it is functionally effective due to the introduction of sensitive outlier indices and locally adaptive and robust statistical criteria. Second, it is computationally efficient because of the use of super-block based spatial data sorting and searching scheme. Our method has been implemented using the C programming language and integrated with the Arc/Info GIS system. The integration leads to a powerful exploratory data analysis tool for checking and analysing anomalous values in a GIS environment. Local outliers can be automatically labeled with our method, subject to some user-defined parameters. Outliers represent anomalous or suspicious values in a statistical sense, which may not necessarily be erroneous values. Instead of being simply discarded, statistical outliers should be investigated further using prior qualitative knowledge or in association with other GIS data layers. |
| |
| Keywords: | |
|
|
