Cognitive Assessment Models with Few Assumptions , and Connections with Nonparametric IRT
Cognitive Assessment Models with Few Assumptions , and Connections with Nonparametric IRT
复制标题
几乎没有假设的认知评估模型以及与非参数IRT的联系
DOI：
--
复制
发表时间：
2001
期刊：
影响因子：
--
通讯作者：
K. Sijtsma
中科院分区：
文献类型：
--
作者：
B. Junker;K. Sijtsma
近年来，随着学习和教学的认知理论变得更加丰富，支持评估的计算方法变得更加强大，使评估真正参考标准的压力越来越大，也就是说，“报告”相对于理论驱动的考生技能，信念和执行特定评估领域任务所需的其他认知特征的学生成绩。认知评估模型通常必须处理比线性排序考生更复杂的目标，或者在低维欧氏空间中部分排序，这就是项目反应理论（IRT）的设计和优化。在本文中，我们考虑了一些可用性和可解释性问题的单策略认知评估模型，该模型描述了一组认知属性之间的随机连接关系进行评估，并在评估中的特定项目或任务的性能。每个考生的属性被编码为存在或不存在，任务被编码为正确或不正确执行。我们考虑的模型做了一些假设潜在属性和任务性能之间的关系超出了一个简单的合取结构：所有相关的任务性能的属性必须存在，以最大限度地提高正确的性能的任务的概率。我们通过例子表明，这些模型可以是敏感的认知属性，即使在数据，设计为良好的配合由Rasch模型，我们考虑了几个随机排序和单调性的属性，提高了模型的可解释性。我们还确定了一些简单的数据摘要，这些数据摘要提供了有关认知属性存在或不存在的信息，当估计模型所需的全部计算能力不可用时。认知评估和NPIRT 1
In recent years, as cognitive theories of learning and instruction have become richer, and computational methods to support assessment have become more powerful, there has been increasing pressure to make assessments truly criterion referenced, that is, to “report” on student achievement relative to theory-driven lists of examinee skills, beliefs and other cognitive features needed to perform tasks in a particular assessment domain. Cognitive assessment models must generally deal with a more complex goal than linearly ordering examinees, or partially ordering them in a low-dimensional Euclidean space, which is what item response theory (IRT) has been designed and optimized to do. In this paper we consider some usability and interpretability issues for single-strategy cognitive assessment models that posit a stochastic conjunctive relationship between a set of cognitive attributes to be assessed, and performance on particular items or tasks in the assessment. The attributes are coded as present or absent in each examinee, and the tasks are coded as performed correctly or incorrectly. The models we consider make few assumptions about the relationship between latent attributes and task performance beyond a simple conjunctive structure: all attributes relevant to task performance must be present to maximize probability of correct performance of the task. We show by example that these models can be sensitive to cognitive attributes even in data that was designed to be well-fit by the Rasch model, and we consider several stochastic ordering and monotonicity properties that enhance the interpretability of the models. We also identify some simple data summaries that are informative about the presence or absence of cognitive attributes, when the full computational power needed to estimate the models is not available. Cognitive Assessment and NPIRT 1