模 糊 無 母 數 統 計 檢 定 及 其在 高 齡 化 社 會 調 查 之 應 用

Abstract

在逐漸高齡化的社會中，關注老人的生活議題並加以分析益顯重要。在研究老人問題時，由於研究對象均曾經歷不同的時空背景與人生閱歷，各個體間存在的差異極大；不同族群的老人對其所慣用語彙的理解與表達亦不盡相同。故若利用傳統的統計分析研究結果，強迫人們採用二元邏輯的方式思考與解釋，可能會導致偏差或錯誤的結論。且傳統的統計檢定方法，往往假定取樣的樣本滿足某個分配，因而導致過多的解釋，影響決策品質。\n 因此，為避免因誤解老人而造成虛耗社會成本，使有限的社會資源得以充分運用，本文於研究老人身心特質與個人期待時，嘗試以模糊理論的軟計算，提出反模糊化轉換。並應用中位數檢定及變異數檢定，建立當統計參數為模糊數或模糊區間時之小樣本無母數模糊統計檢定方法模型。由實證例子分析結果顯示，我們提出的檢定方法，能有效分析模糊樣本的問題，並進而期望能對老人議題的分析和決策有所貢獻，及將此方法運用於其它模糊性議題之研究。

In a gradually aging society, it is important to pay attention to and then analyze elderly people’s life issues. When a study about elderly people is undertaken, the subjects are very inconsistent given their diverse life experience, and various subgroups of subjects have quite different understanding of and way of expressing a vocabulary. Therefore, analyzing study result with conventional statistical analysis method, which forces thinking in a binary logic way, may lead to biased or erroneous conclusion. Furthermore, a conventional statistical test, usually assuming a certain distribution for its samples, may lead to exaggerated explanation, which is detrimental to the quality of a decision.\n So, in order to avoid the waste of our social cost from misunderstanding of the elderly, and to make the most use of our limited social resource, when we investigated the elderly people’s personal characteristics and expectations, we tried to apply the soft calculation of the fuzzy theory, proposed the counter-fuzzy transformation, and, by using the median test and variance test, established a nonparametric fuzzy statistical model for small-sized samples and for parameters of the fuzzy number or fuzzy interval types. The analyses of real-world examples demonstrated that this method of statistical test can analyze the problems of fuzzy samples, and can hopefully contribute to improved analysis and decision making of the elderly people’s issues, and apply this method on the investigation of other fuzzy social issues.\n\nKey words: counter-fuzzy transformation, fuzzy statistical analysis, median test, variance test, aging society.