论文部分内容阅读
认知诊断目前在国内外越来越受到重视,而要真正实现认知诊断,需借助一定的计量学模型即认知诊断模型。国际上,目前已开发近60种认知诊断模型,不同模型又各具特点。本研究重点分析国外研究较成熟的模型——HO-DINA/DINA模型,具体探讨该模型参数估计的实现及模型性能,采用Monte Carlo模拟方法进行,研究发现:(1)本研究对项目参数、属性参数和被试参数估计返真性较好、稳健性较强、HO-DINA模型具有较高的判断率,MCMC算法可行;(2)诊断的属性个数越多,诊断的模式正确率越低,建议实际使用该模型时,诊断的属性个数不宜超过7个;(3)用于诊断的项目数越多,诊断的模式正确率越高,在实际工作中,若要保证有80%以上的模式判准率,则4个属性的至少需20题,5个、6个和7个属性的至少需要40题,8个属性至少需60题。实际运用者应根据实际情况考虑适当的项目数及属性数。
At present, cognitive diagnosis is getting more and more attention both at home and abroad. To truly realize cognitive diagnosis, we need certain metrological model, cognitive diagnosis model. Internationally, nearly 60 cognitive diagnosis models have been developed and different models have their own characteristics. This study focuses on the overseas developed more mature model --HO-DINA / DINA model, discusses the model parameter estimation and model performance, using Monte Carlo simulation method, the study found: (1) This study of the project parameters, The attribute parameters and the parameters of the participants are better than the normal ones, the robustness is strong, and the HO-DINA model has a higher judgment rate and the MCMC algorithm is feasible. (2) The more the number of attributes of the diagnosis, the lower the diagnostic accuracy rate , It is recommended to actually use the model, the number of attributes of the diagnosis should not exceed 7; (3) The more the number of items used for diagnosis, the higher the diagnostic accuracy of the model, in practice, to ensure that more than 80% Of the moderation rate, the four attributes at least 20 questions, 5, 6 and 7 attributes at least 40 questions, 8 attributes at least 60 questions. Actual users should consider the actual number of projects and the number of attributes.