透過您的圖書館登入
IP:52.14.240.178
  • 學位論文

多變量有序資料量化方法及比較分析

Comparison of Scaling Methods for Multivariate Ordinal Data

指導教授 : 許玉雪
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


目前實務界中有序資料的使用日趨頻繁,然而有序資料的統計分析若未經適當處理,貿然採用連續性資料的分析方式,將與許多統計分析方法中連續性的假設前提互相抵觸,因此發展有序資料的量化法是有其必要性,而此多變量有序資料量化法應同時兼顧 (1) 變數值本身應有的順序關係, (2) 變數與變數間的相關性, (3) 量化後資料的連續性。 蕭侑文 (2002) 一文探討多變量有序資料的量化方式,其量化結果可將類別型態的資料轉成連續性的數值資料,惟其在多變量的量化過程中較難保有原始資料的有序結構,因此本研究將以Pölz (1995)及Van de Geer (1992)的量化法為基礎,並結合蕭侑文(2002)提出的隨機化起始值的方式產生不同的多變量有序量化法,其量化過程保留資料的有序性,且使其變數間的相關性達到最大。 為了驗證本研究所提出的多變量有序資料量化法之可行性,分別利用模擬資料和調查所得的多變量有序資料進行實證分析,並就本研究提出的多變量有序量化法與傳統所使用的等距量化法所得之量化結果透過主成份分析、判別分析與Spearman等級相關進行比較分析。 本研究最大的貢獻便是提出多變量有序量化法,使有序類別資料的量化有進一步的突破,量化結果不但保有變數值的順序性、變數間的相關性,且將資料由類別形態提升為連續的資料型態。因此,在未來的各項統計分析上將不受制於連續性的前提假設,有序資料經本文提出的量化法量化後,可利用於各項連續性資料的統計分析方式進行估計、預測及比較。

並列摘要


In practice, ordinal data are usually collected and inappropriately treated as continuous data analyzed with some statistical methods which are only suitable for continuous data. A suitable quantification for ordinal data is necessary that could enable us conduct quantitative analysis with qualitative data. An appropriate quantification method for multivariate ordinal data should simultaneously consider (1) ordinal relationship among levels of the variable, (2) correlation among variables, and (3) continuity of quantified data. Based upon the quantification methods from Pölz (1995) and Van de Geer (1992) this thesis proposes a quantification method for multivariate ordinal data. Both simulation approach and empirical analysis are used to compare the performance of the quantification method with traditional method. The study results show that the quantification method for multivariate ordinal data proposed by this thesis could not only transfer ordinal data into continuous data but also keep the correlation among variables and ordinal relationship among levels of the variable. The multivariate ordinal data rescaled by the methods proposed by this thesis can further be directly analyzed with any statistical methods.

參考文獻


Agresti, A. (1990). categorical data analysis.
Gifi, A. (1990). Nonlinear Multivariate Analysis. New York: Wiley.
Greenacre, J. M. (1984). Theory and applications of correspondence analysis. London: Academic.
Joreskog, G. K., & Moustaki, I. (2001). Factor Analysis of Ordinal Variables: A Comparison of Three Approaches. Multivariate Behavioral Research , 3 (36), pp. 347-387.
Polz, W. (1995). Optimal scaling ordered categories. Computational Statistics , 10, pp. 37-41.

被引用紀錄


彭怡蓁(2011)。有序資料量化之因素分析比較〔碩士論文,國立臺北大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0023-3008201100461600
張佳雯(2012)。線上服務之顧客消費行為及滿意度分析〔碩士論文,國立臺北大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0023-1001201215450400

延伸閱讀