v0.2.1: Post-hoc CLD robustness, covariate role isolation, and inference hardening

Summary

Release focused on inference correctness, robust Compact Letter Display (CLD) calculations, strict separation between experimental treatments and baseline continuous covariates, and flexible block structures.

Changes

• Covariate Role Isolation: Continuous covariates (e.g. plant_density, initial_height) are strictly isolated from post-hoc slicing, factor interactions, and faceting.
• Robust Compact Letter Displays (CLD): Standardized multcompView alignment supporting multi-letter groupings (e.g. ab, bc) and pairwise probability matrices.
• Flexible Blocking Controls: Added explicit block_as = c("auto", "fixed", "random") parameter across suggestion and model fitting workflows.
• Unified Visuals & Reporting: Synchronized post-hoc families, adjustment methods, and estimand definitions across plot_publication() and experiment_report().
• nlme Compatibility: Improved fixed-block model support and variance weighting structures.
• Test Coverage: Added 28 new contract and regression tests in tests/testthat/test-posthoc-contracts.R (188 passing tests total).

Installation

remotes::install_github("PALP31/agriDesignR@v0.2.1")

v0.2.0: Quality, CRAN readiness and stability

Summary

Release focused on package quality, reproducibility, CRAN readiness, and stable statistical workflows. Public API from v0.1.0 remains compatible.

Changes

• Fixed CRAN checks: metadata, imports, namespaces, ASCII source, formula construction, and local artifact handling.
• Added centralized validation for designs, columns, response families, missing values, boundaries, and singular fits.
• Added RNG-safe reproducible layouts and expanded tests covering CRD, RCBD, factorial, split-plot, repeated measures, Gaussian, Poisson, and non-parametric workflows.
• Added CI for release, oldrel-1, and devel R across Ubuntu, macOS, and Windows; coverage and pkgdown workflows.
• Added four vignettes, methodological references, contributor guidance, issue forms, NEWS, and CRAN comments.

Verification

• 38 test blocks pass.
• Clean tarball: R CMD check --as-cran returns 0 errors, 0 warnings, 0 notes.
• pkgdown configuration and Shiny launch path validated.

Installation

remotes::install_github("PALP31/agriDesignR@v0.2.0")

v0.1.0: First Official Release — Decision Engine, Classical ANOVA & Mixed Models

🌱 agriDesignR v0.1.0: First Official Release

Motor Asistido de Decisión, Diagnóstico y Modelos Mixtos para Ciencias Agronómicas y Biológicas

We are excited to announce the initial official release of agriDesignR (v0.1.0)!

agriDesignR is an end-to-end R package designed to bridge the gap between experimental design planning in the field/greenhouse and rigorous statistical modeling for agronomy, plant biotechnology, and experimental biology.

---

🌟 Key Highlights & Features in v0.1.0

1. 🎯 Pre-Experimental Planning & Decision Engine (plan_experiment())

• Rigorous Decision Rules: Provides structured decision criteria to choose between DCA (CRD), DBCA (RCBD), Factorial Designs (1, 2, and 3-way), Split-Plot, Split-Split-Plot, Latin Square, and Repeated Measures.
• Statistical Power & Replications: Automatically calculates the minimum number of blocks/replications required to guarantee residual error degrees of freedom ($df_{\text{error}} \ge 12$).
• Methodological Safeguards: Explicitly alerts researchers against common pitfalls (e.g., misusing Split-Plot for factorials that can be fully randomized, pseudoreplication, or unneeded blocking in homogeneous chambers).

2. 🗺️ 2D Spatial Layout & Planting Map Generator (generate_layout())

• Reproducible Randomization: Generates complete planting schedules with Plot_ID, Block, Whole_Plot, Sub_Plot, and spatial coordinates.
• Phenotyping CSV Templates: Automatically exports ready-to-fill CSV data sheets for direct use during harvest.
• Publication-Quality 2D Croquis: Plots clear 2D bench croquis identifying every greenhouse table, pot position (Pos 01 to Pos 10), and treatment tag.

3. 📊 Dual Statistical Framework: Classical ANOVA vs. Mixed Models (suggest_model(), fit_experiment())

• 🏛️ Classical ANOVA (stats::aov() / stats::lm()): Exact F-tests (Type I, II, III Sums of Squares) and standard Tukey HSD multiple comparisons for balanced designs, fixed greenhouse blocks, and standard thesis tables.
• 🚀 Linear Mixed-Effects Models (lme4::lmer / nlme::lme / glmmTMB): Restricted Maximum Likelihood (REML), Kenward-Roger / Satterthwaite degrees of freedom approximations, and BLUPs for hierarchical split-plots and unbalanced datasets.
• Descomposition in Simple Effects (agri_posthoc(..., by = ...)): Slices 2-way and 3-way factorial interactions to compare genotypes within each environmental level.

4. 💊 Automated Model Remedies (remedy())

• Automatic Box-Cox power transformation ($\lambda$).
• Heteroscedastic variance weighting (nlme::varIdent).
• Generalized Linear Mixed Models (GLMM Gamma, Negative Binomial, Poisson).

5. 📈 Design Efficiency & Biological Rankings (calc_design_efficiency(), agri_ranking())

• Cochran & Cox / Kempthorne Relative Efficiency ($RE%$): Quantifies experimental precision gained by blocking and equivalent replications saved.
• Biological Ranking & Effect Sizes: Computes standardized Cohen's $d$ and percentage gain against control benchmarks.

6. 🎨 Publication Figures & Markdown Reports (plot_publication(), experiment_report())

• Vector graphics (PDF/PNG 300 DPI) styled for Nature and Crop Science with Compact Letter Display (CLD) Tukey letters and colorblind-safe palettes (Forest, Okabe-Ito).
• Complete automated doctoral narrative reports written in Markdown.

7. 🖥️ Interactive Web Dashboard GUI (launch_app())

• Full modern Bootstrap 5 Shiny dashboard featuring 5 interactive modules (Planning, Model Fitting, Diagnostics & Remedies, Publication Figures, and Markdown Report Download).

---

📦 Installation

# From GitHub
if (!requireNamespace("remotes", quietly = TRUE)) install.packages("remotes")
remotes::install_github("PALP31/agriDesignR")

# Launch the interactive GUI
library(agriDesignR)
launch_app()

👤 Author & Affiliation

• Paul Lopez (Estudiante del Doctorado en Biotecnología Vegetal | Profesor Universitario de Aplicaciones Estadísticas, Pontificia Universidad Católica de Chile)
• GitHub: @PALP31