Modified Pena’s Measure for Detecting Influential Observations in Biased Estimators

Abstract

The detection of influential observations is an important aspect of model building in linear regression analysis because of their  unduly large influence on regression coefficients. Different diagnostics measures on identifying these observations have been developed. In this list is the Pena’s statistic diagnostic measure. In this study, Pena’s approach is modified to Liu estimator (LE) and Two-parameter Liu-Ridge estimator (TPE). The performance of the proposed diagnostic for LE and TPE on detecting an influential observation is evaluated with two real life data. Results shows that that the proposed Pena measure performs very effectively in identifying influential observations and agrees with previous studies.