中華心理學刊 民 91，44 卷，2 期，239-251
Chinese Journal of Psychology 2002, Vol.44, No.2, 239-251
王嘉寧（國立台灣大學心理學系）；翁儷禎（國立台灣大學心理學系）
摘要
「探索性因素分析」常為社會科學研究者使用，以作為量表編製和評量測驗效度之用。進行因素分析時，研究者必須仔細考慮變項數目、樣本人數、因素數目、如何估計因素負荷量及轉軸方法，才能獲得穩定可信的結果。此外，資料特性對因素分析結果亦有影響。本研究回顧民國82年至民國88年間，國內心理、教育及管理三個領域中，五種代表性學術期刊使用探索性因素分析的論文，歸納研究者使用因素分析方法時的傾向。針對每一應用因素分析論文，除了討論變項數目、樣本人數、因素數目的決定、如何估計因素負荷量、及轉軸方法等因素分析步驟外，亦針對資料特性中的量尺反應點數與研究者是否考慮得分分配兩項議題進行分析。本研究發現絕大部份研究以題目層次資料進行因素分析，使用的量尺點數大多為五點，取特徵值大於一的個數為因素數目，利用主成份法進行因素負荷量的估計，採正交或斜交轉軸的比例各佔三分之一。文中評論上述方法之優劣，及因素分析應用上宜注意之處。
關鍵詞：探索性因素分析、題目層次因素分析、量尺點數、資料分配
EVALUATING THE USE OF EXPLORATORY FACTOR ANALYSIS IN TAIWAN: 1993-1999
Chia-Ning Wang（Deaprtment of Psychology, National Taiwan University）；Li-Jen Weng（Deaprtment of Psychology, National Taiwan University）
Abstract
Exploratory factor analysis has been widely used by social science researchers for scale development and validity studies. When conducting a factor analysis, the researcher must decide the number of variables to be analyzed, the sample size, the number of factors to retain, the method to estimate initial factor loadings, and the rotation method to be employed. Only when a researcher considers these issues carefully and makes good decisions, can stable and reliable results be obtained. Moreover, the result of applications of factor analysis is also affected by the characteristics of the data. This study reviewed the recent applications of exploratory factor analysis to evaluate the common practice of this frequently used statistical method in Taiwan. Five representative journals ranging from psychology, education, to management between 1993 and 1999 had been reviewed. For every analysis, not only the factor analytic procedures were reviewed, but also the response categories of the scales and the distributions of variables. The results indicated that most studies used item-level variables for factor analysis. Five-point rating scale format was frequently employed. Most studies did not specify the procedures used for deciding number of factors. Among the studies that specified the method to decide number of factors, number of eigenvalues greater than one was the method used most often. Principal component method was popular for estimating initial factor loadings. One third of the studies adopted orthogonal rotations, while another one third adopted oblique rotations. Comments on the frequently used approaches were given. Suggestions for improving the use of factor analysis were also presented.
Keywords：Exploratory factor analysis, Item-level factor analysis, Response categories, Data distribution
