Leung Shing-On, Wu Hui-Ping
Centre for Continuing Education University of Macau, Taipa Macau, China,
J Appl Meas. 2013;14(4):400-13.
Latent Variable Models (LVM) are applied to Rosenberg Self-Esteem Scale (RSES). Parameter estimations automatically give negative signs hence no recoding is necessary for negatively scored items. Bad items can be located through parameter estimate, item characteristic curves and other measures. Two factors are extracted with one on self-esteem and the other on the degree to take moderate views, with the later not often being covered in previous studies. A goodness-of-fit measure based on two-way margins is used but more works are needed. Results show that scaling provided by models with more formal statistical ground correlated highly with conventional method, which may provide justification for usual practice.
潜在变量模型（LVM）被应用于罗森伯格自尊量表（RSES）。参数估计会自动给出负号，因此对于反向计分的项目无需重新编码。可以通过参数估计、项目特征曲线和其他方法找出不良项目。提取了两个因子，一个与自尊有关，另一个与采取适度观点的程度有关，而后者在以往研究中较少涉及。使用了基于双向边际的拟合优度度量，但仍需要更多工作。结果表明，具有更正式统计基础的模型所提供的量表与传统方法高度相关，这可能为常规做法提供了依据。
