项目缺失数据的若干处理技术及其有效性评估——以中国版环境关心量表(CNEP)的应用为例

On Techniques to Handle Item Missing Data and Their Efficiency Assessment :An Case Study on CNEP Scale′s Application

项目缺失数据是定量研究中的一个常见问题,但国内社会学界迄今对此缺乏足够的关注。本文简要介绍了项目缺失数据的成因与机制,并分别回顾了个案删除、单一插补和多重插补3类缺失数据处理技术(MDT)。进一步,利用一个环境社会学的研究示例,对若干MDT的有效性进行了评估,发现缺失数据比例会干扰传统MDT的统计有效性。尽管如此,本文认为,当缺失数据比例较低时(如小于5%),研究者们也可考虑更具实践有效性的一些传统MDT。

Item missing data is a common problem in quantitative research,which so far has received limited attention from Chinese sociology. To start,this paper presents a general introduction to the causes and mechanism of item missing data,and then reviews three types of missing data techniques( MDTs): case deletion,single imputation,and multiple imputation respectively. Further,an example study from environmental sociology is carried out to assess the efficiency of several selected MDTs,which finds that the statistical efficiency of traditional MDTs varies within different rates of missing data. Nevertheless,the paper argues that researchers could also consider some traditional MDTs with more practical efficiency when missing data rate is relatively low( e.g. below 5%).