functional_curves() gains env_smooth and ge_term (with k_ge). env_smooth = TRUE replaces the environment random intercept with a random environment factor-smooth, s(time, env, bs = "fs"), which adjusts for environments that differ in epidemic timing or shape and for treatments that are unbalanced across environments. ge_term = "intercept" or "smooth" adds a genotype-by-environment random intercept or factor-smooth per treatment x environment curve. Defaults keep the previous model. k_env now sets the basis dimension of the environment factor-smooth.
Treatment mean curves average environment factor-smooth and G x E deviations over environments instead of dropping them, because factor-smooths are not centred. The same construction is shared by functional_rate() and diagnose_curves(), so fitted rates match the curves.
Bug fixes
functional_instability() now predicts genotype-by-environment curves with the G x E term, averages out plot-level unit effects, and no longer fails when functional_curves() was fitted with block or unit. It warns when the model has no G x E term, because nFI then mainly reflects each genotype’s disease level.
functional_distances() (curve level) and functional_resistance(group_method = "bootstrap") supply all model variables when predicting curve-level trajectories; the bootstrap in functional_resistance() previously failed when an environment was modelled.
functional_distances(): perm_unit is now honoured. Labels are permuted between permutation units (for example genotypes) and mapped back to their curves, instead of being shuffled between curves; this matters when test_factor is a genotype-level property such as a resistance class. perm_strata is applied at the unit level when strata are constant within units; otherwise whole units are permuted without strata, with a warning.
functional_distances() gains seed (default 1) and, like simulate_dsp_data(seed = 123), no longer resets the user’s random stream: the previous .Random.seed is restored afterwards. The bootstrap seed (boot_seed) is handled the same way.
functional_resistance(instability = ) accepts the list returned by functional_instability(return_curves = TRUE) (it read a non-existent nfi_table element) and joins on the treatment column instead of guessing it.
functional_pca_clusters(choose_k = "silhouette" | "elbow") is implemented; it previously used k = 2 silently. The criterion table is attached as attribute "k_selection".
reconstruct_curves() multiplies back by the column scaling when functional_pca(scale = TRUE) was used, and returns numeric times.
AFSD(): the summary mean shape index now averages the per-focus SIF = rows / cols (it averaged cols / rows).
functional_contrast() averages replicated observations explicitly, with a message, instead of letting approx() average ties silently, and reports when constant extrapolation is used.
functional_summary(): persistence is the total time above threshold, summed over sampling intervals with the profile above the threshold at both ends; gaps below the threshold are no longer counted. Results for profiles without gaps are unchanged.
fit_gradients(): results_table is a data frame with numeric columns; parameters are no longer rounded before back-transformation or before drawing the fitted curves. The documentation no longer claims the models are ranked.
Documentation: FHBWheat$i and $n are the diseased and assessed plant counts used by BPL(); BlastWheat$index_mean is the wheat blast index.
Tests: the CompMuCens() test checked a non-existent element.
r4pde 0.2.2
New functions
Added functional_rate(): Estimates instantaneous rates of plant disease progress (first derivative of trajectory over time) and associated uncertainty from epidemic trajectories fitted by functional_curves(). Supports response and link scale derivatives via GAM linear predictor matrix differences, handles boundary conditions without extrapolation, accounts for zero-epidemic curves and negative rates, and computes key rate phenotypes (r_max, t_r_max, growth duration, cumulative positive growth, and growth windows). Includes S3 methods: print(), summary(), plot(), and augment().
Bug fixes and improvements
functional_curves(): Fixed smooth term label inspection during prediction so that .unit experimental unit and block random effects are correctly identified and supplied to predict.gam().
functional_curves(): Updated the GAM formula to fit independent smooths per treatment by default (trt + s(time, by=trt)), avoiding artifactual drops (“caimento”) at the tails of flat curves. Added the global_smooth parameter to allow users to opt-in to the previous Global-Specific smoothing behavior.
simulate_dsp_data(): Adjusted the amplitudes of the simulated suppression profiles (“Early”, “Late”, “Persistent”) to achieve exactly a 70% reduction in AUDPC compared to the unsprayed control, representing a realistic fungicide efficacy scenario.
Updated _pkgdown.yml to include the newly exported functional PCA and DSP functions in the “Functional Analysis” reference index, resolving site build errors.
functional_pca(): Performs functional principal component analysis on fitted disease progress curves to decompose variation among epidemic trajectories into orthogonal temporal components. Includes plotting and extractor functions.
functional_curves(): Fits genotype-specific epidemic trajectories using GAM, with support for genotype-level covariates.
functional_distances(): Computes pairwise functional distances among fitted curves and performs hierarchical clustering and permutation testing.
functional_resistance(): Calculates Functional Resistance Index (FRI) and Stability-Adjusted Functional Resistance Index (SAFRI) with support for stratified rankings and bootstrap-supported classification.
suggest_k(): Helper function to recommend GAM smoothing parameters (k_smooth, k_trt, k_env, gamma) based on data structure; supports tidy-eval column names and can infer replication from a data frame.
Enhancements
Added support for genotype-level auxiliary covariates (e.g., heading_group) to adjust functional curves and resistance rankings, allowing for better distinction between genetic resistance and phenological escape.
Improved bootstrap methodology for resistance classification with support for stratified group comparisons.
Added functional_instability() for computing normalized functional instability (NFI) of genotype-by-environment epidemic curves, with optional decomposition into spatial and temporal components.
