CMStatistics 2017: Start Registration
View Submission - CMStatistics
Title: Assessing the stability of response styles by using Bayesian item response modeling Authors:  Kensuke Okada - Senshu University (Japan) [presenting]
Daiki Hojo - Senshu University (Japan)
Yusuke Takahashi - Kyoto University (Japan)
Abstract: A response style is the behavioral tendency of selecting particular categories on the rating scale of questionnaire irrespective of the questionnaire's content. Although several previous studies have investigated the stability of response styles, it is inherently difficult to disentangle the substantive trait of interest from the response styles based on self-rating surveys only. We seek to tackle this problem by using measurements through anchoring vignettes, which consist of brief statements describing the situations of hypothetical individuals. The anchoring vignette data allow us to make use of statistical models to adjust the responses of self-rating assessments. In particular, we propose a new Bayesian model for investigating the longitudinal stability of response styles using anchoring vignettes. This is done by extending the multidimensional item response models to incorporate both the substantive trait and the response style. Markov chain Monte Carlo method is used to carry out Bayesian estimation. Information criterion is used to comparatively evaluate the suitability of the time-stable and time-dependent response style models. The proposed method is illustrated through empirical examples. The results show different tendencies between the substantive traits. We discuss the possible explanations and implications of the findings.