透過您的圖書館登入
IP:3.19.31.73
  • 學位論文

大量資料遺失情形下 缺失資料處理方法效用之研究- 以 「 青少年家庭中的親職壓力與親子衝突 」 研究資料為例

The Effect of the Missing Data Techniques when Missing Proportion is Large — A Case Study Using the Parental Stress and Parent-Child Conflict in Adolescents’Families Data

指導教授 : 王鴻龍
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


社會科學研究常需要追蹤並整合兩種以上問卷資料,也常有大量的資料缺漏。本研究旨在探討整合資料有大量資料遺失情形下,使用不同缺失資料處理方法對資料填補的成效。以 1998 年到1999 年「青少年家庭中的親職壓力與親子衝突」研究計畫中的學生、父親、母親整合問卷為實證資料,討論缺失處理方法對青少年家庭親職壓力與親子衝突分析結果的影響。 首先以原始資料中完整學生、家長樣本資料作為基準 (baseline),依據原有的缺失比例,建構 30 組家長問卷的缺失資料集,之後使用四種缺失資料處理方法 (個體刪除法、迴歸填補法、邏輯斯迴歸填補法、蒙第卡羅–馬可夫鏈單一填補法),比較在單因子變異數分析及多元迴歸分析中模型係數與顯著個數的影響。分析結果顯示,迴歸填補法在單因子變異數分析與多元迴歸分析上都是最穩定的缺失資料處理方法。最後我們再以迴歸填補法,對原始資料進行填補與分析。本研究的缺失處理步驟與過程可以做為在面對多類型大量缺失資料型態時,如何進行缺失處理的參考。

並列摘要


Social science research often needs to track or integrate more than two questionnaires. It is frequently occur large number of missing data. In this study, we focus on the imputation effects of various missing data techniques with large portion of missing data. We use the students and parents integration data from 1998 to 1999 of Parental stress and parent-child conflict in adolescents’ families research projects as a empirical data to probe the impact of the missing treatment on the research projects analysis results. First, we use the complete part of the students and parents original data as a baseline. According to the missing proportion from the original data, we constructed the missing data sets from the baseline. Then we compared result of one-way ANOVA and multiple regression analysis among 4 missing treatments, list-wise deletion, regression, logistic regression, and Monte Carlo Markov Chain, with the result based on baseline. The result shows that regression method is the most suitable missing treatments for one-way ANOVA and multiple regression analysis. Finally, we use regression method to impute and analyze the original data. The procedure proposed in our study, to find a suitable missing treatment for large portion of missing data, can be used as a reference for researchers to deal with large missing portion data.

並列關鍵字

Incomplete data Imputation Missing treatment

參考文獻


W. Holmes Finch. (2010). Imputation Methods for Missing Categorical QuestionnaireData : A Comparison of Approaches.Journal of Data Science, 8, 361-378.
王鴻龍、楊孟麗、陳俊如、林定香 (2012)。缺失資料在因素分析上的處理方法之研究。教育科學研究期刊,第五十七卷第一期,29-50。
林欣潔(2012)。缺失資料處理方法對巢狀迴歸分析的影響。國立臺北大學統計學系研究所碩士論文。
Casella, G., and Berger, R. L. (2002). Statistical inference. Duxbury Press.
Kalton, G. and Kasprzyk, D. (1986). The treatment of missing survey data. Survey Methodology, 12(1), 1-16.

延伸閱讀