Conclusions

---

Frontier regression models seek to explain top most or bottom most performers in the data. Many data mining applications can be so conceived. Several potential applications of this type were discussed. Such models are also natural when the data arise from purposeful, goal-directed or managed activities. A test of this characteristic called the normal-like-or-better (NLOB) performance criterion has recently been developed. Using the fitting criterion called maximum performance efficiency (MPE) estimation, the sum of efficiency residuals is minimized. This criterion often reduces to a linear programming model and is therefore straightforward to perform in spreadsheet models with solver capabilities.

---