DEA models with different weight choices and the analysis of the efficiency results

• Main Authors: Chih-ming Wu, 吳志明
• Other Authors: Tsan-hwan Lin
• Format: Others
• Language: zh-TW
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/28910940089544775803

碩士 === 國立高雄第一科技大學 === 運籌管理所 === 95 === DEA (Data Envelopment Analysis) is a common tool for analyzing efficiency and the relationship between multiple inputs and multiple outputs. However, efficiency results vary depending on weight settings. The study investigated different DEA weight models and tried to identify whether certain patterns exist for certain models in certain discriminating criteria. Three criteria are considered. Based on one empirical case and fourteen cases from literature, it was found that common-weight model has higher discriminating power in detecting a small amount of best DMUs (Decision Making Units). CCR model has lower such power. Distribution of efficiency values was defined to check values’ compliance with certain distributions. CCR and non-Archimedean models had higher probability being distributed Normally. Finally, the principal component analysis model had higher variance index for decision makers to differentiate among DMUs’ efficiency values. Decision makers should choose appropriate discriminating criterion and its suggested DEA weight model based on his/her own decision characteristics.