Skip to content

v0.1.2

Choose a tag to compare

View all tags
@PoHsuanLai PoHsuanLai released this 12 Apr 18:29
· 52 commits to master since this release
v0.1.2
a23064d

What's Changed

Fixed

  • Equality constraints (a*V1 + a*V2 syntax) now correctly share a single free parameter. Previously each labelled term got its own independent parameter, silently ignoring the constraint.
  • Labelled first-indicator loadings (e.g. F1 =~ a*V1 + a*V2) are now free, not fixed to 1. Matches lavaan semantics.
  • Observed variable detection for covariance-only models (RL ~~ MDD) no longer fails with "0 observed variables".
  • L-BFGS convergence flag: when the line search exhausts all step sizes, the optimizer is at a stationary point — now correctly reports converged = true.

Added

  • Heywood case warnings: negative variance estimates are detected after every SEM fit and logged at WARN level via the log crate. Surfaces in R, Python, and CLI.
  • New tests for equality constraints (shared params, value propagation, implied covariance, Jacobian) and negative variance detection.

Full Changelog: v0.1.1...v0.1.2

Assets 6
Loading