DBPapers
DOI: 10.5593/SGEM2014/B22/S9.064

THE GOODNESS OF FIT OF LINEAR REGRESSION MODEL IN THE DETERMINATION OF PERMANENT STATIONS’ VELOCITY

J. Bogusz, M. Figurski, A. Klos, A. Araszkiewicz
Wednesday 1 October 2014 by Libadmin2014

References: 14th International Multidisciplinary Scientific GeoConference SGEM 2014, www.sgem.org, SGEM2014 Conference Proceedings, ISBN 978-619-7105-11-7 / ISSN 1314-2704, June 19-25, 2014, Book 2, Vol. 2, 513-520 pp

ABSTRACT
The velocities derived from permanent GNSS stations play more and more significant role nowadays. Therefore their reliable estimation becomes crucial. These velocities and their uncertainties are involved in the kinematic reference frames determination and numerous analyses and interpretations in the geodynamic studies. They are commonly determined with the Least Squares Estimation (LSE) by fitting the trend into changes of topocentric components (North, East and Up). The goodness of fit of such a trend, interpreted as velocity uncertainty, is estimated basing on the differences between the topocentric changes and modelled values. In this research, the changes of daily topocentric components from ASG-EUPOS and selected EPN stations were used. We investigated the goodness of fit of the linear regression line into topocentric components for velocity determination. It was estimated with the coefficient of determination (referred to as R2) and residual error (RE). The R2 values close to 1 prove the well fitted LS line into the time series, while lower values are the effect of unremoved (or improperly removed) outliers, offsets and seasonal components. This clearly demonstrates how determination of the velocity of permanent stations can suffer from the correct data pre-analysis.

Keywords: GPS, ASG-EUPOS, velocity, goodness of fit