Inferring Individual Attributes Using Testlet-Based Visual Analogue Scaling and Beta Copula Diagnostic Classification Models

This paper explores the inference of the latent attributes of respondents using testlet-based visual analogue scaling (VAS), which comprises multiple items ranging from 0% to 100%. Beta copula diagnostic classification models (BCDCMs) are proposed to infer and classify respondents’ latent attributes. The paper also discusses model properties, parameter estimation, goodness of fit, model comparison, and the visual presentation of results. An empirical analysis is conducted on a Symptom Checklist–90 dataset to estimate respondents’ latent psychological and psychiatric symptoms, followed by simulation studies to assess the stability of parameter estimation. The findings indicate that BCDCMs effectively infer latent attributes and derive precise item parameters, as evidenced by the simulation study. This yields valuable diagnostic insights for symptom classification in empirical analyses, suggesting that BCDCMs are viable alternatives to traditional cutoff-score methods for VAS data.