以SEM方法分析精神病學敏感議題資料時採用資料插補法的必要性及適宜性

精神病學之研究資料除了遺漏值問題外，整組資料實屬於多測量多因果的問題。若以傳統的廣義線性模式針對各項測量值分別進行分析，雖能簡化分析結果的解讀，但容易產生結果不一致現象。若能以結構方程式來建構彼此間整體的因果關係，就可有效避免此種結果不一致的現象發生。但傳統結構方程式的使用，除了對整體資料有多變量常態的要求外，更重要的是資料必須是獨立且不得有遺漏值。雖然統計軟體已陸陸續續納入最大期望法的功能，但其適用性(或差異性)卻未呈現。有鑑於此，本研究將針對遺漏值議題，以精神病學之實際研究資料依遺漏值不同嚴重度來進行最大期望法後，再以結構方程式分析並將所獲結果進行比較，進而比較最大期望法的適用性(或差異性)。

關鍵字

遺漏值；最大期望法；結構方程式

並列摘要

Besides the missing data problem, the entire psychiatry data usually is multiple indicators and multiple causes (MIMIC). Using generalized linear models (GLM) to analyze the results for each indicator, the interpretation of the results can be simplified but could end up with inconsistent conclusions. Structural equation modeling (SEM) is an appropriate method to solve this MIMIC problem and avoid the potential inconsistent results. However, SEM require the data to be independently multiple normally distributed and, in addition, no missing data. Although the packages provide EM method, the appropriateness of the analytic results has not been presented. Accordingly, in this study, we are going to compare the appropriateness of using EM methods to various severities of missing data in psychiatry studies based on the results obtained from SEM.

並列關鍵字

Missing data ； Expectation maximization ； Structural equation modeling

參考文獻

Dempster﹐Laird and Rubin(1977)﹐Maximum Likelihood from Incomplete Data via the EM Algorithm,

Daire Hooper, Joseph Coughlan, and Michael R. Mullen (2008), Structural Equation Modelling: Guidelines for Determining Model Fit, Electronic Journal of Business Research Methods, V 6 (n1), p. 53-60.

Little, R.J.A. and Rubin, D.B. (1987), Statistical Analysis with Missing Data, New York: John Wiley & Sons, Inc.

Rubin, D.B. (1987), Multiple Imputation for Nonresponse in Surveys, New York: John Wiley & Sons, Inc.

Google Scholar

Tabachnick, B.G., & Fidell, L.S. (2006). Using multivariate statistics (5th ed.),Needham Heights,MA:Allyn and Bacon.

Google Scholar

國際替代計量

以SEM方法分析精神病學敏感議題資料時採用資料插補法的必要性及適宜性

全文下載

主題瀏覽