Investigation of multicollinearity-influential observations

INVESTIGATION OF MULTICOLLINEARITY-INFLUENTIAL OBSERVATIONS M.Sc. Thesis Ufuk Ekiz GAZİ UNIVERSITY INSTITUTE OF SCIENCE AND TECHNOLOGY June 1997 ABSTRACT Multicollinearity is caused negative influences about the least squares method in regression analysis. For this reason in this thesis influences of multicollinearity are investigated about the least squares method. And then the violence of multicollinearity are investigated by several methods. Variables elimination methods, Robust guess methods and methods of the multicollinearity influential observations are determined. Method of the orthogonal leverage component detenriines multicollinarity-influential observations. We suggested the new method which was similar method of the orthogonal leverage component. This method is compared with the method of orthogonal leverage component in applied. Science code :406.01.01 Key words degression, The Least Squares, Multicollinearity Influential Observation, Orthogonal Leverage Component. Page number :74 Adviser Associated Prof. Dr. Müslim EKNT