Vennebusch, Markus: Singular Value Decomposition and Cluster Analysis as Regression Diagnostics Tools in Geodetic VLBI. - Bonn, 2007. - Dissertation, Rheinische Friedrich-Wilhelms-Universität Bonn.
Online-Ausgabe in bonndoc: https://nbn-resolving.org/urn:nbn:de:hbz:5N-10936
@phdthesis{handle:20.500.11811/2717,
urn: https://nbn-resolving.org/urn:nbn:de:hbz:5N-10936,
author = {{Markus Vennebusch}},
title = {Singular Value Decomposition and Cluster Analysis as Regression Diagnostics Tools in Geodetic VLBI},
school = {Rheinische Friedrich-Wilhelms-Universität Bonn},
year = 2007,
note = {It is well known that high-leverage observations significantly affect the estimation of parameters. So far, mainly redundancy numbers have been used for the detection of single high-leverage observations or of single redundant observations. In this thesis an objective method for the detection of groups of important and less important (and thus redundant) observations is developed. In addition, the parameters which are mainly affected by these groups of observations are identified.
The method proposed in this thesis is based on geometric aspects of adjustment theory and uses the singular value decomposition of the design matrix of an adjustment problem and cluster analysis methods for regression diagnostics.
Although the proposed method can be applied to any geodetic adjustment problem, in this thesis only applications to geodetic very long baseline interferometry (VLBI) are shown. In general, the method is well suited for the detection of (groups of) observations that significantly affect the estimated parameters or that are of negligible impact (and are thus candidates for observations that can be omitted).
In this thesis, at first the theoretical background of the geometrical aspects of geodetic adjustment theory is summarized. Then the singular value decomposition of the design matrix of an adjustment problem is used for the computation of measures of the impact and similarity of observations. Groups of observations with a similar information content are then identified by statistical cluster analysis algorithms. After a short review of geodetic very long baseline interferometry the proposed method is applied to artificial and real single-baseline sessions in order to show the capabilities of the regression diagnostics tool developed in this thesis.},

url = {https://hdl.handle.net/20.500.11811/2717}
}

The following license files are associated with this item:

InCopyright