Quasi-Accurate Detection of Outliers for Correlated Observations

Abstract

Experience with surveying practices has shown that correlated observations are very often encountered, especially in preprocessed observations. Hence, it is not only of theoretical interest, but also of practical interest, to investigate the detection of outliers for correlated observations. The so-called quasi-accurate detection (QUAD) of outliers for correlated observations is developed. The corresponding computation principle and its implementation are investigated in detail. The key of QUAD is how to select the quasi-accurate observations (QAO) reasonably. A new, distinctive sensitivity-analysis based method is proposed for selecting the QAO. For illustrative purposes, an application to global positioning system network adjustment is analyzed. The numerical results demonstrate that more than one outlier can be correctly identified and localized by using the proposed procedure.