In questionnaires with ordinal response categories, observed ratings combine two sources of variation: respondents' substantive evaluations and systematic differences in how they use the response scale. The latter are termed response styles (e.g., midpoint responding, extreme responding, acquiescence). Studying response styles is important because they describe how respondents use the scale, not just what they choose, and simple summaries of the ratings do not capture this information. Most existing approaches either impose a respondent-level style effect shared by all items, or restrict change to pre-specified structures. A key question is whether a respondent's scale usage is stable across the questionnaire or changes across subsets of items; addressing this requires methods that can identify different response-style regimes from the response matrix. We propose a framework that jointly finds clusters of respondents and items, defining respondent-item blocks in which scale-use is homogeneous. Within each block, responses are encoded through cumulative threshold indicators and summarized by an estimated monotone cumulative profile, preserving the category ordering. We assess the approach on simulated rating data to verify that it can recover different scale-use regimes and the associated respondent/item grouping structure.
A Framework for Response-Style Variation in Ordinal Questionnaires / D'Ambrosio, A., Van De Velden, M., Cavicchia, C.. - III:(2026), pp. 88-93. (Statistical Science: From Theory to Applied Research III (SIS-FENStatS 2026) Roma, Dipartimento di Scienze Statistiche, Sapienza Università di Roma 22–25 giugno 2026) [10.1007/978-3-032-30881-8_15].
A Framework for Response-Style Variation in Ordinal Questionnaires
D'Ambrosio Alessia
Primo
Methodology
;Cavicchia CarloUltimo
Supervision
2026
Abstract
In questionnaires with ordinal response categories, observed ratings combine two sources of variation: respondents' substantive evaluations and systematic differences in how they use the response scale. The latter are termed response styles (e.g., midpoint responding, extreme responding, acquiescence). Studying response styles is important because they describe how respondents use the scale, not just what they choose, and simple summaries of the ratings do not capture this information. Most existing approaches either impose a respondent-level style effect shared by all items, or restrict change to pre-specified structures. A key question is whether a respondent's scale usage is stable across the questionnaire or changes across subsets of items; addressing this requires methods that can identify different response-style regimes from the response matrix. We propose a framework that jointly finds clusters of respondents and items, defining respondent-item blocks in which scale-use is homogeneous. Within each block, responses are encoded through cumulative threshold indicators and summarized by an estimated monotone cumulative profile, preserving the category ordering. We assess the approach on simulated rating data to verify that it can recover different scale-use regimes and the associated respondent/item grouping structure.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.


