Abstract
In order to deal with finding the most efficient unit problem, Lam (2015) recently built a new integrated mixed integer linear programming model which is nearly close to the super-efficiency model. The suggested model involves a non-Archimedean epsilon as the lower bound for the input and output weights. Selecting a suitable value for epsilon is a challenging issue in DEA (Data Envelopment Analysis). Lam (2015) suggested a value for epsilon which guarantees the feasibility of his model; however, this paper illustrates that the model may fail to find the most efficient unit due to unsuitable selected value for epsilon. To cope with this issue, a new model is formulated which provides the maximum epsilon value for the model of Lam (2015). The built model guarantees that when epsilon is maximum, then Lam’s model gives exactly one DMU (Decision Making Unit) as the most efficient unit with the maximum discrimination distance from the other DMUs.