《中国家庭追踪调查》中字词测试的信度和效度分析

摘要

大型综合性调查经常包含认知测试部分。基于控制调查成本以及降低数据缺失率的要求，大型调查中的认知测试设计一般较为简洁，在这种情况下，其信度和效度可能受到影响。本文运用经典测试理论和项目反应理论两种方式来分析“中国家庭追踪调查”中字词测试的信度和效度。我们还同时比较了3种计分方法，它们分别是原始分计分法、最难题计分法以及基于项目反应理论的计分法。分析结果显示，“中国家庭追踪调查”中字词测试信度较高，其结构效度和效标效度良好。3种计分方法结果的相关度很高，在分析截面数据时没有实质性的差别。%Many large scale surveys contain cognitive assessment modules. The design of those cognitive tests is often brief in order to minimize test administration cost and maintain acceptable response rates. This paper evaluated the psychometric properties of the literacy test from China Family Panel Studies (CFPS) using both Classical Test Theory (CTT) and Item Response Theory (IRT) approaches. We also compared three different scoring methods:number-right, highest sequence number, and IRT scaled scores. Both CTT and IRT approaches provided positive evidence for reliability and validity of test scores from the CFPS literacy test. Three scoring methods yielded results that were highly consistent with one another.