IRT models for extreme response styles

Kuan Yu JIN, Wen Chung WANG

Research output: Contribution to conferencePapers


Extreme response styles (ERS) refer to a systematic tendency for a person to endorse extreme options (e.g., strongly disagree, strongly agree) on Likert or bipolar rating scale items. Standard IRT models, not considering ERS, will yield biased parameter estimates when ERS exists. In this study we proposed a new class of IRT models to directly account for ERS so that the tendency of ERS is quantified, and the resulting person measures are free from ERS and thus fair. Parameters of the new class of models can be estimated with marginal maximum likelihood estimation methods or Bayesian methods. In this study, the freeware WinBUGS was adopted. In a series of simulations, it was found that the parameters were recovered fairly well; and fitting a standard IRT model yielded substantially biased parameter estimates and slightly underestimated the test reliability. An empirical example of the 2009 International Civic and Citizenship Education Study was provided to illustrate the implication and applications of the new class of models.
Original languageEnglish
Publication statusPublished - Jul 2013


Jin, K.-Y., & Wang W.-C. (2013, July). IRT models for extreme response styles. Paper presented at the 78 Annual Meeting of the Psychometric Society, Arnhem, the Netherlands.


Dive into the research topics of 'IRT models for extreme response styles'. Together they form a unique fingerprint.