Changes in version 0.2.4.9001                      

  - Performance & Stability: Major optimizations to the gradient
    calculation for GCP models. The Jacobian computation now uses a
    sparse matrix approach, leading to significant speedups. The
    $\gamma_1$ correction calculation was corrected by reverting to the
    scalar path, ensuring numerical stability and accuracy.
  - Internal: Added joint_lp_grad and joint_lp_grad_vectorised to the
    output list for external inspection.

                 Changes in version 0.2.4 (2026-04-03)                  

New features

  - bfit_indices() computes per-sample Bayesian fit index vectors
    (BRMSEA, BCFI, BTLI, BNFI), with summary() and print() methods.
    Summary statistics are also available via fitmeasures().
  - compare() compares two or more fitted models side by side, reporting
    marginal log-likelihood, Bayes factors, and DIC, with optional fit
    measures from fitmeasures().
  - diagnostics() computes global and per-parameter convergence and
    approximation-quality diagnostics for fitted models.
  - get_inlavaan_internal() is now exported and documented, providing
    access to the internal list stored in a fitted INLAvaan object.
  - predict() generates predictions for observed data and missing data
    imputation, respecting multilevel structure if present.
  - sampling() draws from the posterior (or prior) SEM generative model,
    returning parameter vectors, latent variables, or observed
    variables.
  - simulate() generates complete replicate datasets from a fitted
    model, useful for simulation-based calibration and posterior
    predictive checks.
  - timing() extracts wall-clock timings for individual computation
    stages of a fitted model.

Minor improvements and fixes

  - Cholesky factorisation of the precision matrix replaces raw solve()
    for covariance and log-determinant calculations.
  - Copula sampling with NORTA (NORmal To Anything) correlation
    adjustment is now the default (samp_copula = TRUE), ensuring
    posterior samples have correct skew-normal marginals and correct
    Pearson correlations.
  - Pre-computed Owen-scrambled Sobol sequences are used by default,
    with fallback to {qrng} for larger sequences. QMC sample size now
    scales with model dimension.
  - Skew-normal fitting now runs in parallel automatically when the
    number of marginals exceeds 120, using all available cores.
  - Small optimisations to the skew-normal volume correction.
  - acfa(), asem(), and agrowth() gain a vb_correction argument.
  - {ggplot2} is now optional; plots fall back to base R graphics when
    it is not installed.
  - inlavaan() gains an sn_fit_ngrid argument to control the number of
    grid points per dimension when fitting skew-normal marginals
    (default 21).
  - inlavaan() now supports sn_fit_sample = TRUE for defined parameters,
    fitting a skew-normal approximation to their posterior marginals
    based on drawn samples.
  - plot() method gains improved visualisation options.
  - priors_for() now supports the [prec] scale qualifier for variance
    parameters (theta, psi), placing the prior on the precision scale
    with automatic Jacobian adjustment.
  - sampling() and simulate() gain a silent argument to suppress
    informational messages.
  - summary() now includes 25th and 75th percentile columns.
  - vcov() now returns the covariance matrix of the lavaan-side
    parameters and supports a type argument for choosing between sample
    and Laplace covariance.

Bug fixes

  - marginal_correction = "shortcut" no longer produces incorrect volume
    corrections.
  - qsnorm_fast() no longer incorrectly handles sign symmetries.

                 Changes in version 0.2.3 (2026-01-28)                  

  - Improved axis scanning, skewness correction, and VB mean correction
    routine.
  - Bug fixes for CRAN.
  - Updated README example.

                 Changes in version 0.2.2 (2026-01-27)                  

  - Under the hood, use lavaan's MVN log-likelihood function to compute
    single- and multi-level log-likelihoods.
  - Added support for multi-level SEM models.
  - Added support for binary data using PML estimator from lavaan. NOTE:
    Ordinal is possible in theory, but the package still lacks proper
    prior support for the thresholds.
  - Added support for missing = "ML" to handle FIML for missing data.

                        Changes in version 0.2.1                        

  - Support for lavaan 0.6-21.
  - Implemented variational Bayes mean correction for posterior
    marginals.
  - Defined parameters are now available, e.g. mediation analysis.
  - Prepare for CRAN release.

                         Changes in version 0.2                         

  - INLAvaan has been rewritten from the ground up specifically for SEM
    models. The new version does not call R-INLA directly, but instead
    uses the core approximation ideas to fit SEM models more
    efficiently.
  - Features are restricted to normal likelihoods only and continuous
    observations for now.
  - Support for most models that lavaan/blavaan can fit, including CFA,
    SEM, and growth curve models.
  - Support for multigroup analysis.
  - Added PPP and DIC model fit indices.
  - Added prior specification for all model parameters.
  - Added support for fixed values and parameter constraints.
  - Initial CRAN submission.

                         Changes in version 0.1                         

  - Used rgeneric functionality of R-INLA to implement a basic SEM
    framework.