Gloucester, Massachusetts.
Department of Psychology, University of Edinburgh, Edinburgh, Scotland.
J Pers. 2019 Aug;87(4):813-826. doi: 10.1111/jopy.12435. Epub 2018 Oct 11.
We tested predictions about the structure and magnitude of method biases in single-source personality trait assessments. We expected a large number of distinct biases that would parallel the observed structure of traits, at both facet and item levels.
We analyzed multimethod ratings on the Estonian NEO Personality Inventory-3 in a sample of 3,214 adults. By subtracting informant ratings from self-reports, we eliminated true score variance and analyzed the size and structure of the residual method biases. We replicated analyses using data (N = 709) from the Czech Revised NEO Personality Inventory.
The magnitude of method biases was consistent with predictions by McCrae (2018, Psychological Assessment). Factor analyses at the facet level showed a clear replication of the normative Five-Factor Model structure in both samples. Item factor analyses within domains showed that facet-level method biases mimicked the facet structure of the instrument.
Method biases apparently reflect implicit personality theory (IPT)-beliefs about how traits and trait indicators covary. We discuss the (collective) accuracy and possible origins of IPT. Because method biases limit the accuracy of single-source assessments, we recommend assessments that combine information from two or more informants.
我们检验了单一来源人格特质评估中方法偏差的结构和大小的预测。我们预计会有大量不同的偏差，这些偏差在特质的层面和项目层面上与观察到的特质结构平行。
我们分析了在 3214 名成年人样本中对爱沙尼亚 NEO 人格量表-3 的多方法评定。通过从自我报告中减去信息源的评定，我们消除了真实分数的方差，并分析了剩余方法偏差的大小和结构。我们使用来自捷克修订版 NEO 人格量表的数据（N=709）重复了分析。
方法偏差的大小与 McCrae（2018，心理评估）的预测一致。在层面水平的因素分析中，两个样本都清楚地复制了规范的五因素模型结构。在域内的项目因素分析中，表明层面水平的方法偏差模拟了工具的层面结构。
方法偏差显然反映了关于特质和特质指标如何共变的内隐人格理论（IPT）-信念。我们讨论了 IPT 的（集体）准确性和可能的起源。由于方法偏差限制了单一来源评估的准确性，我们建议评估应结合来自两个或更多信息源的信息。
