clmstan 0.1.2

Bug fixes / Compatibility

• Updated compatibility with upcoming changes to loo_compare() output structure in the loo package (> 2.9.0), which now returns a data frame instead of a matrix and includes additional diagnostic columns. The package now works with both the previous (matrix) and the new (data frame) output. Thanks to Florence Bockting (loo maintainer) for the reverse-dependency patch (#1).

clmstan 0.1.1

Patch release addressing CRAN maintainer feedback.

Bug fixes

• Suppressed misleading error message during package installation when ‘cmdstanr’ is not available. Stan model compilation is now silently skipped with an informational note instead of an error-like message.

• Removed SystemRequirements: CmdStan from DESCRIPTION. CmdStan is not required to install or check the package; it is only needed at runtime for model fitting.

clmstan 0.1.0

Initial CRAN release.

Features

Link Functions

• 5 standard link functions: logit, probit, cloglog, loglog, cauchit
• 6 flexible link functions with shape parameters:
  • t-link (df)
  • Aranda-Ordaz (lambda)
  • Symmetric Power (r, base)
  • log-gamma (lambda)
  • GEV (xi)
  • AEP (theta1, theta2)

Threshold Structures

• flexible: Standard ordered thresholds (default)
• equidistant: Equal spacing between thresholds
• symmetric: Symmetric thresholds around zero

Prior Specification

• Distribution-based API: prior(normal(0, 2.5), class = "b")
• Support for normal, student_t, cauchy, gamma, and flat priors
• Legacy API via clm_prior() for backward compatibility

Model Comparison

• LOO-CV via loo()
• WAIC via waic()

Diagnostics

• diagnostics() for MCMC convergence checks
• plot() with multiple types: trace, dens, hist, areas, intervals, acf

Prediction

• predict() for class predictions
• fitted() for probability predictions
• posterior_predict() for posterior predictive samples