摘要
追踪研究中测验工具的信度是衡量追踪研究质量的重要指标。传统的信度估计方法不适用于估计追踪研究的测验信度。近年来,研究者提出了四种估计追踪研究的测验信度,包括估计单个时间点的测验信度系数ρw和ρ(Sw),以及估计整个追踪研究的测验信度系数RT和RΛ。本文评述了这四种信度估计方法的数学模型、前提假设及其优缺点。RT和RΛ既可估计追踪研究中单个时间点的测验信度,也可估计追踪研究中整个追踪研究的测验信度,所需要的前提假设较少,推荐同时使用RT和RΛ来估计追踪研究的测验信度。
Test reliability of a longitudinal study is an important index in evaluating the quality of the longitudinal study. Conventional methods that estimate test reliability are not appropriate to estimate the test reliability of a longitudinal study. In recent years, four methods for estimating the test reliability of a longitudinal study were proposed. Pw and p（Sw） coefficients were designed to estimate the test reliability of a longitudinal study at each wave, whereas Rr and R^ were designed to estimate the test reliability of the whole longitudinal study. The mathematical models and assumptions of the four reliability coefficients were briefed and their merits and demerits were pointed out in this article. Rr and RA were recommended because they can be used to estimate the test reliability of a longitudinal study both at each wave and as the whole under a few assumptions.
出处
《心理科学进展》
CSSCI
CSCD
北大核心
2012年第3期467-474,共8页
Advances in Psychological Science
基金
国家自然科学基金项目(30870784)
教育部人文社科重点研究基地项目(11JJD190005)和教育部人文社会科学研究项目基金(09YJCXLX007)资助
关键词
追踪研究
信度
Rr系数
RA系数
longitudinal study
reliability
Rr coefficient
RA coefficient
