LoginSign Up
← Back to all sparks
V

vinereg

ANALYTICS
Velocity0.0

R package vinereg by tnagler — release notes from GitHub.

github.com/tnagler/vinereg ↗

Conditional density and log-likelihood fill out a vine copula regression package.

r-packagecopulasregressionconditional-densitydependency-tracking
◆Current state
vinereg fits D-vine copula-based regression models on top of rvinecopulib and kde1d, in Thomas Nagler's package stack. The January 2025 pair - 0.10.0 and 0.11.0 tagged the same day - adds a pdf() function and then fixes conditional density computation for discrete variables while requiring the newer kde1d. Release notes run to one or two bullets each.
◆Where it's heading
Work has concentrated on evaluation rather than fitting: cll() in 0.9.0, pdf() in 0.10.0, and the discrete-variable correction in 0.11.0 all concern what can be computed from a model already fitted. Releases arrive in same-day pairs, and the notes are terse enough that 0.10.0 reuses 0.9.0's wording verbatim, describing pdf() with cll()'s sentence. Version floors also track the sibling packages - kde1d here, rvinecopulib in 0.8.3.
◆Prediction
Given the shared release rhythm across the stack, the next entry is as likely to be a dependency-driven bump as a new function; the discrete-variable path is the one area these notes show as recently unstable.

◆Recent moves

  1. 1y ago

    Discrete conditional densities fixed; kde1d 1.1.0 required

    Wrong conditional densities for discrete variables is a correctness defect in the package's core output, not a rough edge. Raising the kde1d floor in the same release ties the fix to the sibling package's own version.

    View source ↗
  2. 1y ago

    pdf() added for conditional density

    Adds a second evaluation function after cll(), though the note describes it with cll()'s exact sentence - a copy-paste in the upstream changelog rather than a duplicate feature. Tagged the same day as the fix that follows it.

    View source ↗
  3. 2y ago

    Boost compile flag and a weights error fixed

    A compiler workaround for a removed C++ construct plus a spurious error when weights were passed. Both are the cost of sitting on a compiled dependency stack.

    View source ↗
  4. 2y ago

    cll() computes conditional log-likelihood

    The first of the evaluation-side additions, and the one whose description the later pdf() release reused. It set the direction the 2025 pair continued.

    View source ↗
  5. 4y ago

    vinecopulib floor raised for RcppThread compatibility

    A version floor to keep threading libraries in step, plus clang warnings cleaned up. Invisible to callers, and one of two tags cut that same day.

    View source ↗
  6. 4y ago

    cpit() fixed and external marginals allowed via uscale

    The uscale option lets callers model margins outside the package and pass in uniform data, which widens where vinereg can sit in a pipeline. The cpit() fix - a conditioning step sometimes omitted - is a correctness repair alongside it.

    View source ↗