[M-GI27] Data-driven geosciences

[MGI27-04] Statistical methods for outlier detection

★Invited Papers

*Hideitsu Hino¹ (1.University of Tsukuba)

Keywords:outlier detection

It is inevitable to avoid mixing of outliers in observation and measurement data because of, for example, temporary failure of measurement equipment, influence of noise mixed in communication, or transcription errors from recording paper. There is a possibility that reasoning using data mixed with such outliers may lead to a large error, and detection of outliers automatically and objectively is an essential technique in data analysis. There is also the possibility that irregular data included in ordinary data will lead to scientific discovery. In this presentation, we introduce a method to detect outliers from a statistical point of view.