One of the main issues in DEA modeling is the variables choice. This may have conflictive objectives, like increasing the mean efficiency or maximizing the model ranking capability - a DEA classic fragility. In this paper, we compare four variable selection methods, focused on DMUs sorting. These methods are applied to a real situation of assessing the efficiency of third-party logistics in the activity of newspaper home delivery, in Rio de Janeiro.
DEA; variable selection; ranking
