A COMPARATIVE STUDY OF DIFFERENT OUTLIER DETECTION METHODS IN LINEAR REGRESSION

Outlier detection has been very successful in detecting implausible behaviour points in numerous applications such as voting, network intrusion, weather prediction, etc. In this paper, we compare four diagnostics that are useful in identifying outliers and influential point: the residual plot, leverage, Cook's distance and the modified Thompson tau technique. Using the basic idea underlying identifying outliers and influential points for some data set shows that none is computational efficient.