Computational Strategies and Estimation Performance With Bayesian Semiparametric Item Response Theory Models

IF 1.9 3区心理学 Q2 EDUCATION & EDUCATIONAL RESEARCH

Journal of Educational and Behavioral Statistics Pub Date : 2021-01-27 DOI:10.3102/10769986221136105

S. Paganin, C. Paciorek, Claudia Wehrhahn, Abel Rodríguez, S. Rabe-Hesketh, P. de Valpine

引用次数: 5

Abstract

Item response theory (IRT) models typically rely on a normality assumption for subject-specific latent traits, which is often unrealistic in practice. Semiparametric extensions based on Dirichlet process mixtures (DPMs) offer a more flexible representation of the unknown distribution of the latent trait. However, the use of such models in the IRT literature has been extremely limited, in good part because of the lack of comprehensive studies and accessible software tools. This article provides guidance for practitioners on semiparametric IRT models and their implementation. In particular, we rely on NIMBLE, a flexible software system for hierarchical models that enables the use of DPMs. We highlight efficient sampling strategies for model estimation and compare inferential results under parametric and semiparametric models.

查看原文本刊更多论文

贝叶斯半参数项目反应理论模型的计算策略和估计性能

项目反应理论（IRT）模型通常依赖于对特定主题潜在特征的正态性假设，这在实践中往往是不现实的。基于狄利克雷过程混合物（DPM）的半参数扩展为潜在特征的未知分布提供了更灵活的表示。然而，IRT文献中对此类模型的使用极为有限，这在很大程度上是因为缺乏全面的研究和可访问的软件工具。本文为从业者提供了关于半参数IRT模型及其实现的指导。特别是，我们依赖NIMBLE，这是一个用于分层模型的灵活软件系统，可以使用DPM。我们强调了模型估计的有效采样策略，并比较了参数和半参数模型下的推断结果。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Journal of Educational and Behavioral Statistics Multiple-

CiteScore

4.40

自引率

4.20%

发文量

期刊介绍： Journal of Educational and Behavioral Statistics, sponsored jointly by the American Educational Research Association and the American Statistical Association, publishes articles that are original and provide methods that are useful to those studying problems and issues in educational or behavioral research. Typical papers introduce new methods of analysis. Critical reviews of current practice, tutorial presentations of less well known methods, and novel applications of already-known methods are also of interest. Papers discussing statistical techniques without specific educational or behavioral interest or focusing on substantive results without developing new statistical methods or models or making novel use of existing methods have lower priority. Simulation studies, either to demonstrate properties of an existing method or to compare several existing methods (without providing a new method), also have low priority. The Journal of Educational and Behavioral Statistics provides an outlet for papers that are original and provide methods that are useful to those studying problems and issues in educational or behavioral research. Typical papers introduce new methods of analysis, provide properties of these methods, and an example of use in education or behavioral research. Critical reviews of current practice, tutorial presentations of less well known methods, and novel applications of already-known methods are also sometimes accepted. Papers discussing statistical techniques without specific educational or behavioral interest or focusing on substantive results without developing new statistical methods or models or making novel use of existing methods have lower priority. Simulation studies, either to demonstrate properties of an existing method or to compare several existing methods (without providing a new method), also have low priority.