背景:项目反应理论在国外的应用非常广泛,但在国内的非认知测量中却较少应用。因此需要探讨项目反应理论在非认知测量中是否可行,了解哪些模型最为适合。目的:拟以情感能力量表为例,探讨自陈量表分析中如何选择最佳的项目反应理论模型,从模型与数据的拟合,模型提供的信息量等方面比较不同模型的有效性。设计:采用心理测验法,在同一批测试数据中比较不同的心理测量学模型。单位:山东莱芜职业技术学院和辽宁师范大学心理学系。对象:观察对象为2004-05山东师范大学和莱芜市第十七中学在校的617名大学生(男311人,女306人)和564名高中生(男283人,女281人),共1181人参加了测试。方法:测量工具为9因素情感能力量表,量表从冲动控制、共感能力、坚持性、人际亲密、社交能力、情绪调节、情绪稳定、责任感、自信9个维度测量情绪与社会能力,项目为5级评分。对9个分量表分别用单参数、双参数和三参数Logistic模型和等级反应模型进行分析,并进行拟合度检验(类似χ2检验),再对这三种模型的测量精度(信度)进行比较。主要观察指标:9个维度在采用单参数、双参数和三参数Logistic模型时的模型拟合指标对数似然比,拟合的χ2平均值,及模型的平均信息量。结果:发放问卷1220份,删除无效问卷后得回答规范问卷1181份。单参数、双参数和三参数Logistic模型三者相比,双参数模型的对数似然比是最小的,均方根残差大于2的项目数也最少,提供的信息量大于单参数模型而不低于三参数模型。因此在采用2级评分模型时双参数模型是最好的。但双参数Logistic模型的测量精度低于多级项目反应模型。结论:自评量表若是2级评分资料,则采用双参数Logistic模型分析是可行的,符合项目反应理论的标准,而单参数和三参数Logistic模型则不理想。当自评量表为多级评分时,其测量精度要好于2级评分,如果合并选择项,则测量精度下降。 Background: The project response theory is widely used in foreign countries, but it is rarely used in non-cognitive measurement in China. Therefore, it is necessary to explore whether the project response theory is feasible in non-cognitive measurement to know which models are most suitable. OBJECTIVE: To explore how to select the best project response theory model from the self-report scale analysis by taking the emotion ability scale as an example, and to compare the validity of different models from the fitting of the model and the data and the amount of information provided by the model. Design: Using psychometric methods to compare different psychometric models in the same batch of test data. Unit: Shandong Laiwu Vocational and Technical College and Liaoning Normal University Department of Psychology. PARTICIPANTS: Subjects were 617 college students (311 males and 306 females) and 564 high school students (283 males and 281 females) from Shandong Normal University and the 17th Middle School in Laiwu City in 2004-05, for a total of 1181 People participated in the test. Methods: The measurement tools were 9-factor emotional ability scale. The scale measures emotional and social ability from 9 dimensions including impulse control, common sense ability, persistence, interpersonal intimacy, social ability, emotional regulation, emotional stability, responsibility and self-confidence. 5 ratings. The nine subscales were analyzed by one-parameter, two-parameter and three-parameter logistic model and rank response model, respectively, and the fitness test (similar to χ2 test) was conducted. The accuracy of these three models Compare MAIN OUTCOME MEASURES: Log-likelihood ratio of model fitting index, mean χ2 of fitting, and average information amount of model in nine dimensions in single parameter, two parameter and three parameter Logistic model. Results: 1220 questionnaires were distributed, and 1181 questionnaires were answered after the invalid questionnaires were deleted. The logarithmic likelihood ratio of the two-parameter model is the smallest compared to the single-parameter, two-parameter and three-parameter Logistic models, and the number of items with the root mean square residual greater than 2 is the least, providing more information than the single parameter model Not less than the three-parameter model. Therefore, the two-parameter model is best when using a 2-level scoring model. However, the measurement accuracy of the two-parameter Logistic model is lower than that of the multi-level project response model. Conclusion: If the self-rating scale is 2-level score data, it is feasible to use the two-parameter Logistic model analysis, which is in line with the standard of the project response theory. However, the one-parameter and three-parameter Logistic models are not ideal. When the self-rating scale is a multi-level score, the measurement accuracy is better than the 2-level score, and if the selection items are combined, the measurement accuracy decreases.