Lathrop Quinn N, Cheng Ying
Pearson, Portland, USA.
University of Notre Dame, Notre Dame, USA.
Psychometrika. 2017 Mar;82(1):245-263. doi: 10.1007/s11336-016-9513-1. Epub 2016 Oct 3.
Item cloning is increasingly used to generate slight differences in tasks for use in psychological experiments and educational assessments. This paper investigates the psychometric issues that arise when item cloning introduces variation into the difficulty parameters of the item clones. Four models are proposed and evaluated in simulation studies with conditions representing possible types of variation due to item cloning. Depending on the model specified, unaccounted variance in the item clone difficulties propagates to other parameters in the model, causing specific and predictable patterns of bias. Person parameters are largely unaffected by the choice of model, but for inferences related to the item parameters, the choice is critical and can even be leveraged to identify problematic item cloning.
项目克隆越来越多地用于在心理实验和教育评估中生成任务的细微差异。本文研究了项目克隆在项目克隆的难度参数中引入变异时出现的心理测量学问题。提出了四种模型，并在模拟研究中进行评估，模拟条件代表了由于项目克隆可能产生的变异类型。根据指定的模型，项目克隆难度中未解释的方差会传播到模型中的其他参数，从而导致特定且可预测的偏差模式。人参数在很大程度上不受模型选择的影响，但对于与项目参数相关的推断，模型选择至关重要，甚至可以利用它来识别有问题的项目克隆。
