Diagnostics measures for identification of outliers in multiple regression

In multiple linear regression; ordinary least squares analysis does not give satisfactory and consistent results in the presence of linear depency among predictors (multicollinearity) and existence of outliers in data.In the literature, several biased estimator have been proposed as alternatives to the least squares estimator in the presence of multicollinearity to mitigate the effect of multicollinearity in the analysis. A class that includes a part of biased estimator has been proposed by Lee and Birch (1988). After that, it was shown that Liu and generalized Liu estimators can also be included in this class by Topçubaşı (2001). In this study, it is shown that modified ridge and modified Liu estimators can also be included in this class.Little work has been done on the use of diagnostic measures for biased estimator. In this thesis, diagnostic measures are also defined for this class of biased estimator by using classical diagnostics measures given for assessing the influence of observation on least squares regression results.