Selecting window width in data envelopment analysis using a method based on analytical hierarchical process

Abstract

Operational research is an application of mathematical, statistical, and computing techniques to solve management decision making problems. Data envelopment analysis (DEA) is a nonparametric method in operational research and economics for the approximation of production frontiers. It is used to empirically evaluate a productive efficiency of decision making units (DMUs). Nonparametric approaches have the benefit of not presuming a peculiar functional pattern for the frontier; however, they do not provide a general relationship (equation) relating output and input. Window analysis is one of the extended DEA methods. The objective of this thesis is to formulate a methodology for the selection of window width in DEA window analysis. Based on this, the problem is defined as selecting a window width that is narrow. In order to accomplish this aim it will be a necessity to develop an abstract model in order to cover what can be fulfilled by empirical application and simulation studies. The proposed methodology is integrated into the framework of a conceptual model for the selection of window width that is structured by combining two multi-criteria decision making techniques. A new approach used in this thesis, has proven successful, and the result agreed with the preconceived idea. A hypothesis is tested, and it is proven that the newly developed model is independent of the empirical approach. A result has shown that the selected window width is robust to optimum efficiency assessment. This new method is superior than the empirical approach in performance evaluation.