以WHOQOL-BREF探討傳統與模糊量表間的一致性與測驗恆等性

Abstract

基於過去傳統測量方法與人類主觀認知在性質上的落差，過去二十年中，主張將模糊理論納入量表測量的研究廣泛的出現在各種領域。而針對這些研究，過去學者發現它們所測量到的模糊數具有提供個案諮商訊息、表徵複雜心理構念、驗證特殊理論等傳統測量工具無法提供的功能，而將這些模糊數轉換成明確實數後，其信效度也優於傳統的測量工具。但是，這些研究並無法回答使用模糊量表的研究成果，是否能和過去使用傳統測量工具的研究互相參照的問題。因此本研究即利用WHOQOL-BREF台灣版做為工具，採受試者內設計，比較受試者在隸屬度加權量尺(FPWS)、模糊圖形量表(FGRS)、模糊部份給分法(FPCS)、Chen與 Hwang所提出模糊計分法(CH)等四種模糊計分方式上與傳統量表的一致性和測驗恆等性(ME/I)。本研究利用404位受試者的資料進行分析，分別利用Cronbach α係數、組內相關係數(ICC)、確認式因素分析進行內部一致性信度、一致性與ME/I的分析。本研究結果顯示FPWS、FGRS、FPCS的內部一致性信度較傳統的李克特式量尺(LS)好，而在範疇層次的分析亦顯示它們與LS間具有良好的一致性。而在測驗恆等性方面，進入分析的FPWS、FGRS、FPCS皆接受了因素負荷量恆等性的模型，且方法間的差異並未在觀察分數上顯示出實質的影響。故總結來說，研究者認為這三種模糊計分與傳統測量間的結果是可以互相參照的，但是受限於研究工具以及樣本特性，未來的研究可以針對不同領域的量表、不同的模糊計分、不同背景的族群做進一步探討。

In past two decades, researchers have proposed to combine fuzzy theory into measurement in various areas, basing on the difference between properties of human cognition and traditional measure methods. According to their studies’ results, the fuzzy numbers collected by fuzzy scale can provide more consulting information , verify special theory and represent complicated constructs; even if we transform these fuzzy numbers to crisp numbers, the results collected by fuzzy scale are superior to those by traditional measurement in both validity and reliability. However, the comparability of results between fuzzy scale and traditional method is yet to confirm by previous studies. Hence, in present study, researcher use WHOQOL-BREF Taiwan version as a instrument to test agreement and Measurement Equivalence /Invariance (ME/I) between fuzzy scale weighted by membership (FPWS), fuzzy partial credit scaling(FPCS), fuzzy graphic scale(FGRS), fuzzy scale proposed by Chen and Hwang(CH) and traditional scale in repeated measurement experimental design, with a set of data from 404 subjects. Reliability, agreement and ME/I were applied by using cronbach’s alpha Coefficient, Intraclass Correlation Coefficient and confirmatory factor analysis respectively. The results indicate that the reliability of FPCS, FPWS, FGRS are superior to traditional Likert Scale(LS) , and all four fuzzy scales show almost perfect agreement with LS at domain level. In ME/I analyses, FPCS, FPWS and FGRS accepted the model that constrained equal factor loading and show no substantial difference on manifest variables with the traditional scale. To sum up, combine the results of agreement and ME/I analysis, researcher suggest that the results collected by FPCS, FPWS, FGRS and traditional scales are comparable, but limited by study’s instrument and characters of subjects. Further investigation of agreement and ME/I properties between fuzzy and traditional scales would lie on other sets of population and research area.