A unified model for detecting efficient and inefficient outliers in data envelopment analysis

Data quality is critical to a successful data envelopment analysis (DEA) study. Outlier detection not only identifies suspicious data points and thus prevents the drawing of erroneous conclusions, but also can lead to the discovery of unexpected knowledge. This study develops a unified model to identify outliers in DEA studies by examining how they effect on the boundaries of a data set. The proposed model is mostly consistent with primary efficient influential data detection studies. It is also able to detect inefficient outliers, which is yet unsolved, based on the ground of DEA assumptions.