Patch
- Improved convergence assessment and Hessian-rescue behavior,
including more robust selection of the best candidate fit.
- Treats lower-bound solutions for softplus covariance-diagonal
parameters as admissible rather than automatically classifying them as
convergence failures.
- Hardened parameter-bound detection, nudging, and bound
relaxation.
Patch
- Revised the model-implied mean vector and covariance matrix for
distal-outcome models so that distal outcomes are predicted by latent
effect sizes rather than observed effect-size estimates.
- Added support for treating covariates as either fixed definition
variables or stochastic variables modeled jointly with effect sizes and
distal outcomes.
- Added Monte Carlo method confidence intervals.
Patch
Patch