基于IRT理论传统纸笔测验与计算机自适应测验结果对比分析

发布时间：2018-05-17 09:27

本文选题：CAT + IRT　；参考：《贵州师范大学》2014年硕士论文

【摘要】：基于项目反应理论的计算机自适应测验（Computerized AdaptiveTest,CAT）是一种比传统测验方式更加快速有效的测验方式。目前，对于CAT的研究多集中在CAT系统的建立、CAT相关算法研究与优化、CAT与其他测验理论的结合三个方面。罕有文章对CAT在测验中的准确性及测验效率的进行普适研究，尤其是研究在理想状况下CAT与传统纸笔测验方式的差异与特点。本文以EPQ量表当中N量表为基础，首先使用真实数据进行模型拟合与参数估计，研究在实际测验中，CAT的可行性以及其测验结果与传统纸笔测验的相关与差异；而后利用Monte-Carlo生成模拟数据，对基于IRT的传统纸笔测验及CAT结果进行分析，排除各种干扰因素之后，进一步研究CAT与传统纸笔测验在理想状况下，其理论与模型的固有特点，得到以下结果：（1）CAT效率的提升，是以测验准确性的有控制的牺牲为条件的，其实质是在测验效率与测验准确率做平衡，每一个项目的减少，都会造成测验准确性的牺牲，如何有计划的筛选项目，是CAT首要的问题。（2）题库项目的增加，对于整个被试群体来说，会使传统纸笔测验和CAT的测验准确性增加，传统纸笔测验准确性的提升，同时表现在测验标准误的降低上，而在CAT中，由于本文限定了终止条件的标准误，最终被试能力估计值的标准误并没有因为测验准确性的提升而改变，这说明传统纸笔测验或者CAT所获得的被试能力估计的标准误，无法表明其测验的准确性。（3）对于不同能力水平的被试，题库项目的增加，可能会提升其测验准确性，也可能会提升其测验效率。
[Abstract]:Computerized Adaptive Test (CAT) based on item response theory is a more rapid and effective test method than traditional methods. At present, the research of CAT mainly focuses on the establishment of CAT system and the combination of cat and other test theories. Few articles have studied the accuracy and efficiency of CAT in testing, especially the differences and characteristics between CAT and traditional paper and pen test methods under ideal conditions. On the basis of N scale in EPQ scale, the model fitting and parameter estimation of real data are used to study the feasibility of cat in the actual test and the correlation and difference between the results of the test and the traditional paper and pen test. Then we use Monte-Carlo to generate simulation data, analyze the traditional paper pen test and CAT result based on IRT, remove all kinds of interference factors, and further study the inherent characteristics of the theory and model of CAT and traditional paper pen test under ideal condition. The following results were obtained: The improvement of cat efficiency is conditioned by the controlled sacrifice of test accuracy, which is essentially a balance between test efficiency and test accuracy, and the reduction of each item will result in the sacrifice of test accuracy. How to plan the selection of items, is the primary issue of CAT. 2) for the whole group, the increase of item bank will increase the accuracy of traditional paper and pen test and CAT test, improve the accuracy of traditional paper and pen test, and decrease the error of test standard, while in CAT, the accuracy of traditional paper and pen test will be improved, while in CAT, the accuracy of traditional paper and pen test will be improved. Because the standard error of the termination condition is limited in this paper, the standard error of the final test ability estimate is not changed by the improvement of the test accuracy, which indicates that the standard error of the traditional paper and pen test or the test ability estimation obtained by CAT. The accuracy of the test cannot be demonstrated. 3) for the subjects with different ability levels, the increase of test bank items may improve the accuracy and efficiency of the test.
【学位授予单位】：贵州师范大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：B842

【参考文献】