李克特式量表廣為社會科學所使用,其傳統計分方式為等距的整數數值,但這種計分方式尚存有疑慮與待商榷之處,例如等距變數的假設過於牽強、要求填答者二分法的選擇語意等。 鑑於李克特式量表的廣為使用,以及部份的研究者嘗試以模糊語意變數的計分方法,改進傳統的計分方式,但有關傳統計分和模糊語意變數計分對信度的影響,這方面的文獻並不多見,值得進一步深入探討。 基於此,本研究從探討傳統計分和模糊語意變數計分角度,以資料模擬的研究方法,探討在試題數、語意模糊度、內部一效性等因數的影響下,傳統計分和模糊語意變數計分的信度差異。經由資料模擬顯示,模糊語意變數計分的α係數顯著地高於傳統計分的α係數,這方面的資料模擬結果也指出模糊語意變數計分的適用性和可行性。最後,研究者根據資料模擬結果,提出未來實務和研究上的建議。
The purpose of this research is to investigate the properties of reliability for fuzzy linguistic variables. Likert Scale of questionnaire has been widely used by social science researcher. The score calculating of this scale is transforming linguistic terms into interval variables of integers. And its method of responding questions is to select just one linguistic term which fits the attitude best. But the above scaling exists some doubts and it isn't appropriate to explain the exact psychological state. There are some literatures which use fuzzy linguistic variables to assess the attitude. Most of these literatures show that fuzzy linguistic variables are proper method of score calculating for Likert Scale. But these literatures don't discuss the reliability of fuzzy linguistic variables in Likert Scale questionnaires. So, it is important and necessary to investigate the reliability of fuzzy linguistic variables. For the purpose of discussing the reliability, the factors of items numbers, fuzziness of linguistic terms, and internal consistency of items are manipulated. According to the results of data simulation, we could realize that the a coefficient of fuzzy linguistic variable is significantly higher than that of traditional score calculating. Finally, based upon the findings of this study, some recommendations for further research are suggested.