Abstract:
Identifying extreme values in a data set is one of the preliminary and necessary analysis in a nonparametric study when each decision making unit is essential in defining the efficient frontier. Many studies have proposed methodologies for identifying outliers in an unknown distribution setting and contradictions have appeared down the road. This study offers an empirical illustration of the method of identifying outliers proposed by Simar in 2003 for frontier order-m but in a probabilistic setting and a comparison between the results using different models of efficiency. The technique is used to identify influential observations in a data set containing Romanian universities. Sensitivity analysis related to the different values of the parameter α revealed important modifications in the shape of the robust frontier in case of outliers. The results can be used to create an initial homogeneous data set for further efficiency analysis.
