因子分析在问卷调查中信度效度评价的应用

The Application of Factor Analysis in Reliability and Validity Assessment

进行问卷式调查研究的统计分析前，应评价问卷的信度和效度，以确保结果的准确性与可靠性。用单次调查资料评价问卷信度的θ系数法及Ω系数法可克服常用的重复测量法的不足。利用因子分析评价问卷的结构效度是最理想的效度评价方法。本文以“健康行为问卷”为例，介绍因子分析在信度和效度评价方面的应用。分析发现，使用克朗巴哈α系数法低估了整份问卷的内部一致性，而基于因子分析的θ及Ω系数法则较准确。除第２分项外，问卷的５个分项各只有１个公因子，且有较满意的因子载荷量（≥０．４），说明具有良好的结构效度。第２分项的两个公因子可认为分别反映了欲测概念，以及问卷题目的设计方式。

It is essential to assess the reliability and validity of questionnaire,to ensure the quality of surveillance data and outcomes of statistical analysis afterward.Based on the information gathered from test for once,theta (θ) coefficient and omega (Ω) coefficient are able to make up the weakness of testretest stability,in terms of the reliability assessment.It's viewed as the best protocol of employing factor analysis in construct validity assessment.A psychometric study on the 'Health Behavior Questionnaire' as an example,the application of factor analysis was introduced.It was found that Cronbach α coefficient under estimated the internal consistency.On the other hand,θ and Ω coefficients,established upon factor analysis,provided closer estimates to the true reliability.Findings showed the unifactorial structure with moderate to high loading for each sigle subscale,except the second pat of scale.The extracted two factors of the second sub scale were considered as a measured conceptual factor and a method artifact,the factor caused by different response sets toward positive and negative items.