Morales-Vives Fabia, Ferrando Pere J, Hernández-Dorado Ana
Psychology Department, Research Center for Behavior Assessment, Universitat Rovira i Virgili, Tarragona, Spain.
J Gen Psychol. 2025 Jul-Sep;152(3):375-402. doi: 10.1080/00221309.2024.2404398. Epub 2024 Sep 18.
Most IRT applications in personality assume that the measured trait is a bipolar dimension, normally distributed in the population. These assumptions, however, could be questionable for maladaptive, (quasi) pathological traits that still fall in the normal range. This study focuses on one such trait, Callousness, and uses two different instruments and samples to determine whether there is a basis for modeling it as a unipolar trait instead of a bipolar one. More specifically, the following community samples were used, recruited in several Spanish high schools: a) 719 adolescents (13-19 years old, 55.8% girls), b) 681 adolescents (13-19 years old, 44.9% girls). Callousness was assessed with the Inventory of Callous-unemotional traits and Antisocial behavior in the first sample and with the Inventory of Callous Unemotional traits in the second sample. We compared the outcomes of fitting the Graded-Response model (a bipolar-trait model) and the Log-Logistic model (a unipolar trait model) in these community samples and found that they differed considerably at the scoring level. In terms of accuracy, the conditional reliability functions had opposite patterns: it was maximum at high levels in the Graded-Response model and at low levels in the Log-Logistic model. In terms of validity, the models showed different results regarding the prediction of indirect aggressiveness and non-planning impulsiveness.
人格领域的大多数IRT应用都假定所测量的特质是一个双极维度，在总体中呈正态分布。然而，对于仍处于正常范围内的适应不良的（准）病理性特质而言，这些假设可能存在问题。本研究聚焦于这样一种特质——冷酷无情，并使用两种不同的工具和样本，以确定是否有依据将其建模为单极特质而非双极特质。更具体地说，我们使用了从西班牙几所高中招募的以下社区样本：a）719名青少年（13 - 19岁，55.8%为女生），b）681名青少年（13 - 19岁，44.9%为女生）。在第一个样本中，使用冷酷无情特质与反社会行为量表来评估冷酷无情；在第二个样本中，则使用冷酷无情特质量表。我们比较了在这些社区样本中拟合等级反应模型（一种双极特质模型）和对数逻辑斯蒂模型（一种单极特质模型）的结果，发现在评分水平上它们有很大差异。在准确性方面，条件可靠性函数呈现出相反的模式：在等级反应模型中，高水平时条件可靠性函数达到最大值；而在对数逻辑斯蒂模型中，低水平时达到最大值。在效度方面，这两个模型在预测间接攻击性和非计划性冲动方面显示出不同的结果。
