Exploring aberrant responses using group response functions

UNCG Author/Contributor (non-UNCG co-authors, if there are any, appear on document)
James Robert Davis (Creator)
The University of North Carolina at Greensboro (UNCG )
Web Site: http://library.uncg.edu/
Micheline Chalhoub-Deville

Abstract: The purpose of this study was to investigate the utility of empirical group response functions (GRFs) for contextualizing aberrant responses in educational test data. A GRF illustrates the functional relationship between expected item proportion correct score (classical p-value) and the item difficulty scale, given a specified examinee subgroup and item set or subset. Lack of consistency between empirical and expected (modeled) GRFs may suggest aberrance related to subgroup characteristics, item characteristics, or their interactions. Under relatively ideal simulation conditions, the GRF approach explored in this study appeared to provide an accurate and sensitive representation of subgroup-level aberrance over different item subsets and aberrance types. A demonstration of the approach using real data from a K-12 testing context uncovered a substantive interaction. GRF patterns by item passage type (informational versus literary) were qualitatively different, but only for examinees designated as English Learners (EL). This result suggests that EL students used different response strategies for different content types, which has implications for both validity and fairness issues. Overall, results provide support that the GRF approach represents a meaningful contribution toward the call for more comprehensive, explanatory person fit methodologies.

Additional Information

Language: English
Date: 2020
Educational Testing, Graphical Fit Approaches, Group Response Functions, Person-Fit Analysis, Psychometrics, Subgroup-level aberrance
Educational tests and measurements
Goodness-of-fit tests

Email this document to