Analyzing Response Styles With Mixture Item Response Theory Model and Partial Credit Model in Response Styles: A Simulation Study

• In Event: Psychometrics and Educational Measurement Poster Session 1

Abstract

Response styles in non-cognitive assessments are problematic since they affect the
validity of scale scores and thus the trait estimates are biased. This study focuses on
extreme response styles (ERS), and surveys popular methods in analyzing EMS. A
simulation study is utilized to compare the performance of three models (i.e., simple
PCM, mixture PCM, and PCMRS) with regards to parameter recovery. This study
found that the mixture PCM consistently recovered person trait better than the
simple PCM but worse than the PCMRS, and it recovered threshold parameters even
worse than the simple PCM. Some explanations of poor performance of the mixture
PCM is provided. Limitations and further research on extending mixture IRT to
analyze response styles are also discussed.

Author

• Zebing Wu, City University of New York Graduate Center