가중치에 대한 제약 및 분석 표본 수에 따른 DEA 효율성과 참조집단의 변화에 대한 실험연구

다국어 초록 (Multilingual Abstract)

DEA is value free and does not require specification or knowledge of prior weights or prices for inputs or outputs. If there is some informations on value judgements on virtual weights, we can improve the quality of DEA results. This study investigates the relationships between DEA efficiency scores, weight restrictions and data sizes by simulation experiments. Also this paper interests in the behaviour of reference sets and which factors influence the change of reference sets. Some results of this study are as follows. Overestimation in efficiency score arises when using unbounded DEA, but this phenomenon can be reduced to some extent with additional restrictions in virtual weights and increased sample sizes. After eliminating the effects of the assumed real efficiency scores, efficiency scores systematically changed by restriction in weights and sample sizes. The number of reference sets is reduced as sample sizes decreased, but the degree of reduction rate is smaller in sample with restrictions.