Exploring latent groups of response styles on a self-concept scale based on the mixed partial credit model

This study aimed to identify latent groups based on response styles to a self-report assessment and to explore the effects of these response styles on the estimation of respondent parameters and the correlations between sub-scales. The data from the 2013 Korea Education Longitudinal Study, specifically from the survey of third-year middle school students on the self-concept measure, were used for the analyses. The main findings are as follows. First, applying a mixed partial credit model to explore latent groups according to response styles on each sub-scale, it was found that, in the social self-concept scale, there emerged two groups: a non-extreme response style group and an extreme response style group. Second, in the physical and academic self-concept scales, two non-extreme response style groups with difference preferences of selecting options and an extreme response style group were identified. Third, in the social, physical, and academic self-concept scales, male students preferred extreme response styles more than female students. Lastly, for the group using extreme response styles, the respondent parameter was overestimated when response styles were not accounted for, while for the group using non-extreme response styles, the respondent parameter was underestimated when the response styles had not been controlled. After controlling for response styles, the correlations between sub-scales decreased.