Man Kaiwen, Harring Jeffrey R
University of Alabama, Tuscaloosa, AL, USA.
University of Maryland, College Park, MD, USA.
Educ Psychol Meas. 2021 Jun;81(3):441-465. doi: 10.1177/0013164420968630. Epub 2020 Oct 31.
Many approaches have been proposed to jointly analyze item responses and response times to understand behavioral differences between normally and aberrantly behaved test-takers. Biometric information, such as data from eye trackers, can be used to better identify these deviant testing behaviors in addition to more conventional data types. Given this context, this study demonstrates the application of a new method for multiple-group analysis that concurrently models item responses, response times, and visual fixation counts collected from an eye-tracker. It is hypothesized that differences in behavioral patterns between normally behaved test-takers and those who have different levels of preknowledge about the test items will manifest in latent characteristics of the different data types. A Bayesian estimation scheme is used to fit the proposed model to experimental data and the results are discussed.
为了理解正常表现和异常表现的考生之间的行为差异，人们提出了许多方法来联合分析项目反应和反应时间。除了更传统的数据类型外，生物特征信息（如来自眼动仪的数据）可用于更好地识别这些异常的测试行为。在此背景下，本研究展示了一种用于多组分析的新方法的应用，该方法同时对从眼动仪收集的项目反应、反应时间和视觉注视次数进行建模。据推测，正常表现的考生与那些对测试项目有不同程度的先验知识的考生之间的行为模式差异将体现在不同数据类型的潜在特征中。采用贝叶斯估计方案将所提出的模型拟合到实验数据中，并对结果进行了讨论。
