Item-Focussed Trees for the Detection of Differential Item Functioning in Partial Credit Models

TL;DR

This paper introduces an item-focused tree method for detecting differential item functioning in partial credit models, especially for ordered response categories, providing visual and interpretable results.

Contribution

It presents a novel tree-based approach to identify DIF in partial credit models, extending DIF detection to ordered response data.

Findings

The method effectively detects DIF items in simulations.

Visual trees clearly show variables inducing DIF.

Application on real data demonstrates practical utility.

Abstract

Various methods to detect differential item functioning (DIF) in item response models are available. However, most of the methods assume that the responses are binary, for ordered response categories available methods are scarce. In the present paper DIF in the widely used partial credit model is investigated. An item-focussed tree is proposed that allows to detect DIF-items, which might affect the performance of the partial credit model. The method uses tree methodology yielding a tree for each item that is detected as DIF-item. The resulting trees show which variables induce DIF and in which way. The visualization as trees makes the results easily accessible. The method is compared to an alternative approach, simulations demonstrate the performance of the method and an application illustrates how it works for real data.