Scholarlink

<p>The detection of differential item function (DIF) is a crucial task in item response theory modeling. In recent years, machine learning (ML) techniques are increasingly used for this task, for example, using model-based recursive partitioning (MOB) techniques. For example, Rasch trees are a combination of MOB and Rasch models. In this paper, we propose an alternative ML technique for DIF detection in Rasch models, called exceptional model mining in Rasch models (RaschEMM). While Rasch trees and RaschEMM share an algorithmic base and the goal to identify covariate-based subgroups, they have different goals. Rasch trees aim at identifying a “global model of the data”, whereas RaschEMM aims at identifying exceptional subgroups. After a general and comprehensive presentation of RaschEMM, we present the results from two simulation studies and provide an applied example.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Mining exceptional Rasch models

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年1月13日

查看全文
  • Ch. Kiefer,
  • M.-A. Sengewald

摘要

The detection of differential item function (DIF) is a crucial task in item response theory modeling. In recent years, machine learning (ML) techniques are increasingly used for this task, for example, using model-based recursive partitioning (MOB) techniques. For example, Rasch trees are a combination of MOB and Rasch models. In this paper, we propose an alternative ML technique for DIF detection in Rasch models, called exceptional model mining in Rasch models (RaschEMM). While Rasch trees and RaschEMM share an algorithmic base and the goal to identify covariate-based subgroups, they have different goals. Rasch trees aim at identifying a “global model of the data”, whereas RaschEMM aims at identifying exceptional subgroups. After a general and comprehensive presentation of RaschEMM, we present the results from two simulation studies and provide an applied example.