Individual differences in response scale use: Mixed Rasch modelling of responses to NEO-FFI items

Abstract

Mixed Rasch modelling allows the existence of sub-groups who differ in their use of questionnaire response scales to be investigated. This is relevant to personality measurement, as individual differences in scale use are a source of error in calculating trait scores. The objectives of the analysis reported here were to investigate the use of the mixed Rasch model on personality (NEO-FFI) data and to examine the issue of response set contributions to scale-level NEO correlations. Modelling led to two-class solutions for Neuroticism, Agreeableness and Conscientiousness, which could be interpreted in terms of respondents showing a preference for the extremes or the middle of the scale. Correlations between all pairs of class assignments were positive and highly significant, demonstrating consistency in individual response preference across traits. Plots of item difficulty parameters suggested that both groups interpreted the items similarly. ‘Extreme responders’ had significantly higher scores on Extraversion and Conscientiousness, and females and younger people were more likely than males and older people to show a preference for extreme responding. For Extraversion and Openness, model fitting gave more complex results, but a two-class solution was considered to also be the most appropriate for Extraversion. The results of this study confirm previous findings on the existence of distinct response classes in self-report data and on personality correlates of individual differences in response scale use and also allow an estimate of response set effects on scale-level correlations.