Package index • fancyfx

Effect plots

Every effect curve is paired with a rug of the raw data stacked directly above it, so a bend can be read against how much data supports it. GAMs go through gratia, everything else through marginaleffects.

effect_estimates(): Extract a variable's effect from a model as a tidy data frame
permutation_importance(): Permutation importance for a fitted model
plotEffects(): Plot a predictor's effect with a rug of the raw data above it
plotImportance(): Plot permutation importance
plotRugs(): Create rug plots representing distribution of the raw data
plotSmooths(): Extract and plot smooths from a GAM (deprecated)

Performance and calibration

Whether the model is right, and whether its probabilities mean what they say.

calc_deviance(): Deviance of a set of predictions
calibration_estimates(): Are the model's predicted probabilities honest?
held_out(): Evaluate predictions you already have
plotCalibration(): Plot a calibration curve with a rug of where predictions fall
plotROC(): Plot a ROC curve
plotThreshold(): Plot classification metrics against the decision threshold
print(<fancyfx_held_out>): Print a held_out object
threshold_metrics(): Threshold-dependent classification metrics across every cutoff

Extrapolation and novelty

Where a prediction is being made outside the data it was fitted on. MESS and the sorting-bias measures say how far outside.

ensemble_summary(): Summarise an ensemble of projection rasters
mess(): Multivariate environmental similarity surface
niche_equivalency(): Test an observed overlap against a null of interchangeable occurrences
niche_overlap(): How much do two predicted distributions overlap?
plot(<fancyfx_equivalency>): Plot a niche equivalency test against its null
plotExtrapolation(): Map where a projection leaves the conditions the model was fitted under
plotUncertainty(): Map the disagreement between ensemble members
spatial_sorting_bias(): Spatial sorting bias in a train/test split

Composition and style

Arranging panels, binning dense scatter, and the palette and theme that keep a set of figures looking like one set.

combinePlots(): Combine multiple effect plots for simultaneous display
comparePlots(): Compare the same effect across several models
fancyfx_palette(): Colours for effects split into several curves
hex_bin(): Aggregate spatial values into hexagonal bins
plotHexbin(): Map values aggregated into hexagonal bins
theme_fancyfx(): A publication-ready theme for effect plots
thin_points(): Thin points so that no cell holds more than a few