两种不同的项目反应理论的比较研究

发布时间：2018-05-03 06:31

本文选题：项目反应理论 + 基本假设　；参考：《西南大学》2017年硕士论文

【摘要】：目前,随着测量学家的深入研究,项目反应理论已日趋成熟,其中以洛德、汉布尔顿的项目反应理论最具盛名。迄今为止,IRT还是有许多理论问题尚待解决。比如,人们容易把单维性假设等同于局部独立性假设,单维性假设并不满足实际测验中的要求,局部独立性假设为什么要加上“局部”两字?模型假设只是为了拟合测验得到的数据吗?IRT仅仅是一种经验化的理论吗?若不是,那它的理论基础又是什么?本文将会围绕这些问题展开讨论。杜文久于2014年出版了《高等项目反应理论》一书,该书以概率论为工具,对IRT的基本假设、反应模型的理论基础都作出了详细讨论,并对IRT中的许多定理、公式进行了论证。在《高等项目反应理论》中,杜提出了与洛德、汉布尔顿所不同的三条基本假设,因而得到不同的结论,主要表现在:(1)在保序假设之下,能力参数可以被定义,每一个被赋予数值的被试的能力参数具有不变性,被试在各测验项目上的百分位不变,即能力参数不变。但在洛德、汉布尔顿的IRT中能力参数只是一个抽象的概念,仅是一个描述潜在特质的术语,并没有量化;(2)包括二参数模型、三参数模型、二级评分模型和多级评分模型在内的所有项目反应模型,都可采用节点的方法演绎推理得到对应的概率分布函数,从而项目反应模型就拥有了理论基础。而在洛德、汉布尔顿的IRT中,所有的项目反应模型都只为拟合数据,缺乏理论基础;(3)当项目参数与能力参数都未知需要同时估计时,其能力参数估计值依概率收敛于真值。而洛德、汉布尔顿的项目反应理论不具这样的性质;(4)杜运用演绎的方式对IRT中的定理和公式作了推理论证,得出IRT未来的发展趋势是转向演绎化的道路。从总体来看,洛德、汉布尔顿的项目反应理论仍表现为经验化,缺乏一定的逻辑基础。
[Abstract]:At present, with the in-depth study of the surveying scientists, the theory of project response has matured, in which the theory of project response of Lord, Han Burton is the most famous. So far, there are many theoretical problems to be solved in IRT. For example, it is easy to equate the single dimension hypothesis to the assumption of local independence, and the single dimension hypothesis is not satisfied with the actual measurement. Why should the hypothesis of local independence add to the "local" character? The model hypothesis is only to fit the data obtained by the test? Is IRT just an empirical theory? If not, what is its theoretical basis? This article will be discussed around these issues. Duwen has published < advanced project for a long time in 2014. In the book of reaction theory, the book, with the theory of probability, discusses the basic hypothesis of IRT, the theoretical basis of the reaction model, and demonstrates many theorems and formulas in IRT. In the higher project reaction theory, Du puts forward three basic assumptions different from Lord, Han Burton, and thus get different conclusions. The main performance is as follows: (1) under the preorder hypothesis, the capacity parameters can be defined, the capacity parameters of each given value are invariable, the percentile of the test on each test item is invariable, that is, the capacity parameter is constant. But the energy parameter in IRT of LOD and Han Burton is only an abstract concept, only a description of the potential. The idiosyncratic terms are not quantified; (2) all project response models, including the two parameter model, the three parameter model, the two grade scoring model and the multilevel scoring model, can be deductive and deductive to the corresponding probability distribution function by the node method, and the project response model has the theoretical basis. In LOD, the IRT of the Han Burton. All of the project response models are only data fitting and lack of theoretical basis. (3) when the parameter and capability parameters of the project are unknown, the estimation of its capability parameters converges to the true value in probability according to the probability. And Lord, the theory of the project reaction of Han Burton does not have such properties; (4) the Du Yunyong deductive way to the theorem in IRT and The formula makes a reasoning and argumentation, and concludes that the future trend of IRT is the road to deduction. In general, the theory of project response of LOD and Han Burton is still experienced and lacks a certain logical basis.

【学位授予单位】：西南大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O211

【参考文献】