Date of Award

Spring 5-12-2023

Degree Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Educational Policy Studies

First Advisor

Dr. Hongli Li

Second Advisor

Dr. Audrey Leroux

Third Advisor

Dr. T. Chris Oshima

Fourth Advisor

Dr. Yinying Wang

Abstract

The Differential Test Functioning (DTF) statistic, with the Item Parameter Replication (IPR) procedure, can measure Differential Item Functioning (DIF) within the Differential Functioning of Items and Tests (DFIT) framework for Item Response Theory (IRT) models. However, it comes with many practical costs and theoretical assumptions. In some reasonably anticipated circumstances, the DTF statistic cannot be evaluated easily, and DFIT analysis consequentially remains beyond the scope of impacted IRT models. A straightforward, diagnostic statistic would add value to typical IRT model fitting. It was hypothesized that a statistic based on Mahalanobis distances and standard errors of an IRT model could perform as a reliable flag for likely DIF. To test this hypothesis, a Monte Carlo simulation study compared the performance of the traditional DTF measure to the new statistic. Although easy to calculate, the statistic proved unproductive in flagging models with DIF present. Related performance analysis and recommendations were provided.

DOI

https://doi.org/10.57709/35505275

File Upload Confirmation

1

Share

COinS