Package index • pigauto

One-call entry point

The main user-facing function. Runs the full pipeline (preprocess_traits → build_phylo_graph → fit_baseline → fit_pigauto → predict → evaluate) and returns a completed data frame.

impute(): Impute missing phylogenetic traits (convenience wrapper)

Experimental analysis-aware MI for fixed effects

The analysis model is declared before imputations are drawn. Initial support is one incomplete continuous covariate under MAR with lm, binomial-logit glm, or one-random-intercept lmer. Fixed effects only; the backend passed its package-level fixed-effect gate and remains an experimental, deliberately narrow interface.

multi_impute_analysis(): Analysis-aware multiple imputation for narrow regression models
with_imputations(): Fit a downstream model on every imputed dataset
pool_mi(): Pool downstream model fits across multiple imputations (Rubin's rules)

Stochastic prediction diagnostics

Conformal-width, Brownian/MC-dropout, PMM, and posterior-tree draws. These paths failed or were outside the downstream inferential gate and must not be used as analysis-aware multiple imputations.

multi_impute(): Generate experimental stochastic completion datasets
multi_impute_trees(): Posterior-tree prediction sensitivity

Pipeline — fine-grained control

Individual steps of the impute() pipeline, exposed for benchmarking and custom workflows.

preprocess_traits(): Preprocess trait data: align to tree, encode into latent space
build_phylo_graph(): Build a phylogenetic graph representation from a tree
make_missing_splits(): Split cells into train/val/test for imputation evaluation
mask_missing(): Create an observed/missing mask matrix
fit_baseline(): Fit the phylogenetic baseline
fit_pigauto(): Fit a pigauto model for trait imputation
predict(<pigauto_fit>): Impute missing traits using a fitted pigauto model
evaluate(): Evaluate a fitted pigauto model on its test set
evaluate_imputation(): Evaluate imputation performance against known values

Active imputation (sampling-design guidance)

Per-candidate-observation expected uncertainty reduction across every currently-missing cell, using closed-form BM (Sherman-Morrison) variance reduction for continuous / count / ordinal / proportion / zi_count magnitude and label-propagation entropy reduction for binary / categorical / zi_count gate.

suggest_next_observation(): Suggest which cell to observe next to maximise imputation precision

Covariate data helpers

Optional helpers for assembling environmental covariate matrices from public data sources. Both are designed to plug into impute(..., covariates = X).

pull_gbif_centroids(): Fetch species range-centroid covariates from GBIF
pull_worldclim_per_species(): Fetch per-species bioclim covariates from WorldClim v2.1

Benchmarks and cross-validation

simulate_benchmark(): Run a simulation benchmark for pigauto
simulate_non_bm(): Simulate non-BM trait data for benchmarking
cross_validate(): k-fold cross-validation for pigauto trait imputation
compare_methods(): Compare BM baseline and pigauto methods across replicates

Reporting and plotting

pigauto_report(): Generate an HTML benchmark report from a pigauto fit
plot(<pigauto_fit>): Plot diagnostics for a fitted pigauto model
plot(<pigauto_pred>): Plot predictions from a pigauto model
plot(<pigauto_benchmark>): Plot a pigauto benchmark
plot_comparison(): Forest-plot style comparison of benchmark results
plot_history_gg(): Plot training history (ggplot2, deprecated)
plot_uncertainty(): Plot uncertainty ribbons for imputed trait values
summary(<pigauto_fit>): Summary method for pigauto_fit objects
calibration_df(): Compute calibration data for probability predictions
confusion_matrix(): Compute a confusion matrix for categorical or binary predictions

I/O helpers

read_traits(): Read trait data from a CSV file or data frame
read_tree(): Read a phylogenetic tree from a file
save_pigauto(): Save a fitted pigauto model
load_pigauto(): Load a saved pigauto model

Bundled data

avonet300: AVONET morphological and ecological trait data for 300 bird species
tree300: Example bird phylogeny for the 300 species in avonet300
trees300: 50 posterior phylogenies for the 300 species in avonet300
avonet_full: Full AVONET morphological and ecological trait data for 9,993 bird species
tree_full: Example bird phylogeny for the species in avonet_full
ctmax_sim: Simulated multi-observation-per-species CTmax data