2026-06-16

OrdinalRegressionViz

Project · R package

OrdinalRegressionViz

Visual tools that connect model parameters to latent evidence, response thresholds, and uncertainty.

Latent signal and noise distributions with ordered response thresholds
Latent distributions and the thresholds that convert evidence into response categories.

Question

How can an ordinal model become inspectable?

Ordinal regression parameters are easier to discuss when the latent evidence, response criteria, and resulting categories can be seen together.

What I built

Plots and measures.

The package draws latent distributions, ROC curves, posterior predictive checks, and Bayesian forest plots. It also computes d′, criteria, scale ratios, and AUC.

Key design decision

Keep the thresholds visible.

The visual language connects the coefficients to the response rule instead of treating the fitted model as a black box.

Artifact

Open the package README

Limitations

A plot is not a design.

A clear figure cannot repair weak data, an unexamined coding choice, or an unsupported causal claim.

Linked from