CMStatistics 2022: Start Registration
View Submission - CMStatistics
B1276
Title: Modelling ``don't know'' responses in multi Item Latent Trait Models Authors:  Maria Iannario - University of Naples Federico II (Italy) [presenting]
Abstract: The focus is on a general mixed item response theory (IRT) framework that allows for the differentiation of respondents with respect to the type of processes underlying item responses. The study analyses the presence of response styles in answering and don't know responses. The latter is considered qualitatively different and is modelled using an additional latent variable that captures respondents' willingness to respond. Naive approaches, based on the evaluation of the option as incorrect answer or missing value, can lead to biased measures of the latent trait(s). The mixed model is studied with simulation studies and applied to an empirical example concerning the effect of financial technology on financial inclusion.