pkgdown/assets/favicon.html

Skip to contents

Changelog

Source: NEWS.md

OptiDesign 0.1.0

Initial release

New design constructors

  • Added prep_famoptg(): constructs repeated-check block designs with flexible replication, supporting augmented, partially replicated (p-rep), and RCBD-type repeated-check layouts. The p-rep constraint — no replicated treatment appears twice in the same block — is enforced by construction.

  • Added alpha_rc_stream(): constructs fixed-grid alpha row-column stream designs with repeated checks. The field is treated as a global traversal stream partitioned into contiguous replicate segments and incomplete blocks. Block-size constraints are expressed as total block size (checks + entries) via min_block_size and max_block_size.

New evaluation functions

  • Added evaluate_famoptg_efficiency(): standalone mixed-model efficiency evaluator for designs produced by prep_famoptg(). Model contains Block + Row + Column random effects (sigma_b2). Computes A-criterion, D-criterion, and CDmean. Fully decoupled from construction.

  • Added evaluate_alpha_efficiency(): standalone mixed-model efficiency evaluator for designs produced by alpha_rc_stream(). Model contains Rep + IBlock(Rep) + Row + Column random effects (sigma_rep2, sigma_ib2). Computes A-criterion, D-criterion, and CDmean. Fully decoupled from construction.

New optimisation functions

  • Added optimize_famoptg(): Random Restart (RS) optimiser for prep_famoptg() designs. Searches across n_restarts independent randomisations and returns the design with the best criterion value. Supports A, D, both, and CDmean criteria. RS is used exclusively because the p-rep constraint is preserved by construction at every call.

  • Added optimize_alpha_rc(): criterion-driven optimiser for alpha_rc_stream() designs with three search strategies:

    • RS (Random Restart): independent randomisations, return best.
    • SA (Simulated Annealing): iterative entry-permutation swaps with temperature-governed acceptance, repeated across multiple restarts.
    • GA (Genetic Algorithm): population of entry permutations evolved via Order Crossover (OX1), swap mutation, and elitism. Supports A, D, both, and CDmean criteria. All methods preserve structural constraints by construction.

Optimality criteria

Both evaluation and optimisation functions support:

  • A-criterion: mean pairwise contrast variance (fixed effects) or mean prediction error variance / PEV (random effects). Lower is better.
  • D-criterion: geometric mean of contrast covariance eigenvalues (fixed effects only). Lower is better.
  • CDmean: mean coefficient of determination for genomic breeding value prediction, CDmean = 1 - mean_PEV / sigma_g2 (Rincent et al. 2012). Higher is better. Requires random treatment effects and a genomic prediction model (prediction_type = "GBLUP" or "PBLUP").

Integrity checking

Both optimisers implement a four-point integrity checking strategy: post-construction validation, re-check before storing as best, final check before return, and emergency fallback if no valid design is found across all iterations.

Internal architecture

  • Added alpha_rc_helpers.R: internal helper functions shared across all six exported functions — sparse incidence matrix construction, AR1 precision matrices, the mixed model solver, the Hutchinson stochastic trace estimator (Hutchinson 1990) for scalable criterion approximation in large designs, Chebyshev neighbourhood enumeration, and the dispersion scoring function.

  • Construction, evaluation, and optimisation follow a single-responsibility architecture: each step is a separate function that can be called independently or chained.

Supported residual structures

All evaluation and optimisation functions support:

  • "IID": independent residuals.
  • "AR1": row-only AR1 autocorrelation.
  • "AR1xAR1": separable row × column AR1 autocorrelation.

Grouping and dispersion

Both design families support:

  • Family-based adjacency scoring (cluster_source = "Family").
  • GRM-based clustering (cluster_source = "GRM").
  • Pedigree-based clustering (cluster_source = "A").
  • Optional post-layout genetic dispersion optimisation (use_dispersion = TRUE) using swap-based local search to reduce spatial clustering of related entries.

Package infrastructure

  • Added package documentation (OptiDesign-package.R) covering both design families, typical workflows, function overview table, and key references.
  • Added OptiDesign_example_data: synthetic dataset with treatment vectors, field dimensions, relationship matrices (GRM, A, K), and ready-to-use argument lists for both design families.
  • Added pkgdown site configuration (_pkgdown.yml) with reference sections organised by design family.
  • Added GitHub Actions workflows for R CMD CHECK and pkgdown site deployment.
  • Added unit tests (testthat >= 3.0.0).