Detection of Outliers in GPS Measurements by Using Functional-Data Analysis

Abstract

The identification of outliers in global positioning system (GPS) observations—to compare equipment, positioning methods, or working conditions—has traditionally been performed using univariate or multivariate statistics. However, these methods have certain drawbacks when processing data collected by GPS receivers. Such data can be more suitably handled as observations at discrete points of a smooth stochastic process and, consequently, other statistical approaches to the analysis of functional data may prove more suitable. We analyzed the applicability of the concept of functional depth to the identification of outliers in GPS observations. The proposed method was applied to 12 series of GPS receiver data collected in an open space and in similar signal reception conditions. The results obtained adapted better to the expected results, given the signal-reception conditions, than those obtained by the classical statistical approaches used by other writers to compare GPS observations.