BIRM-RS: model and software
Response styles are systematic tendencies to favor particular regions of a rating scale, regardless of what an item asks. They distort means, variances and correlations and can bias group comparisons.
Visual analogue scales produce bounded continuous data, which standard models for Likert-type items are not designed for. Together with Tanja Lischetzke and Mirka Henninger, I extended the Beta Item Response Model (Noel & Dauvier, 2007) so that it captures extreme, midpoint and acquiescent responding at the same time.
Simulation
The model is estimated in Stan. A simulation study on an HPC cluster (Slurm) showed that the parameters are recovered well when a questionnaire contains enough items, including reverse-scored ones. The simulation tools are available as the R package birmrssim, citable via doi.org/10.5281/zenodo.20305322.
Interactive apps
Both apps run in the browser with shinylive, no server needed:
- BIRM-RS: how person, item and response-style parameters shape the distribution of responses
- Zero- and one-inflated BIRM-RS: a proof of concept for a proposed extension to responses at exactly 0 and 100
Papers
- Vollbracht, D., Lischetzke, T., & Henninger, M. (2026). Detecting response styles in visual analogue scales. PsyArXiv. doi.org/10.31234/osf.io/xteku_v2
- Vollbracht, D., Müller, M., Lischetzke, T., & Henninger, M. (2026). Extreme response style generalizes across Likert-type and visual analogue scales. PsyArXiv. doi.org/10.31234/osf.io/zevb7_v1