Donnellan M Brent, Ackerman Robert A, Brecheen Courtney
a Department of Psychology , Texas A & M University.
b School of Behavioral and Brain Sciences, The University of Texas at Dallas.
J Pers Assess. 2016;98(2):169-77. doi: 10.1080/00223891.2015.1058268. Epub 2015 Jul 20.
Although the Rosenberg Self-Esteem Scale (RSES) is the most widely used measure of global self-esteem in the literature, there are ongoing disagreements about its factor structure. This methodological debate informs how the measure should be used in substantive research. Using a sample of 1,127 college students, we test the overall fit of previously specified models for the RSES, including a newly proposed bifactor solution (McKay, Boduszek, & Harvey, 2014 ). We extend previous work by evaluating how various latent factors from these structural models are related to a set of criterion variables frequently studied in the self-esteem literature. A strict unidimensional model poorly fit the data, whereas models that accounted for correlations between negatively and positively keyed items tended to fit better. However, global factors from viable structural models had similar levels of association with criterion variables and with the pattern of results obtained with a composite global self-esteem variable calculated from observed scores. Thus, we did not find compelling evidence that different structural models had substantive implications, thereby reducing (but not eliminating) concerns about the integrity of the self-esteem literature based on overall composite scores for the RSES.
尽管罗森伯格自尊量表（RSES）是文献中使用最广泛的整体自尊测量工具，但对于其因子结构仍存在持续的分歧。这场方法学辩论影响着该测量工具在实证研究中的使用方式。我们以1127名大学生为样本，检验了先前为RSES指定的模型的整体拟合度，包括新提出的双因子解决方案（麦凯、博杜泽克和哈维，2014年）。我们通过评估这些结构模型中的各种潜在因子与自尊文献中经常研究的一组标准变量之间的关系，扩展了先前的研究。一个严格的单维模型对数据的拟合度很差，而考虑了正负计分项目之间相关性的模型往往拟合得更好。然而，可行结构模型中的整体因子与标准变量以及与根据观察分数计算的综合整体自尊变量所获得的结果模式具有相似的关联水平。因此，我们没有找到令人信服的证据表明不同的结构模型具有实质性影响，从而减少了（但并未消除）基于RSES整体综合分数对自尊文献完整性的担忧。
