Package index • ggBoostedTrees

Extract — pull a data frame out of a fit

Start here. Each extractor takes a fitted model and returns a tidy data frame with a documented column contract, so what a figure shows can be read, tested and reshaped before anything is drawn. gg_boost_error() needs a fit grown with cv.flag = TRUE; boostmtree records the error path and the optimal iteration only when cross-validation ran.

gg_boost_error(): Boosting error trajectory data object
gg_boost_path(): Boosting parameter path data object
gg_boost_trajectory(): Subject trajectory data object
gg_boost_vimp(): Variable importance data object
gg_boost_effect(): Partial and marginal effect data object
boostmlr_partial(): Partial effects from a BoostMLR fit
gg_boost_calibration(): Calibration over follow-up data object

Render — draw the extracted object

The autoplot() method for each extracted class, with plot() as its alias. Rendering reads only the column contract above, which is what lets a new modelling backend be added by writing extractor methods and nothing else.

autoplot(<gg_boost_error>) plot(<gg_boost_error>): Plot a gg_boost_error object
autoplot(<gg_boost_path>) plot(<gg_boost_path>): Plot a gg_boost_path object
autoplot(<gg_boost_trajectory>) plot(<gg_boost_trajectory>): Plot a gg_boost_trajectory object
autoplot(<gg_boost_vimp>) plot(<gg_boost_vimp>): Plot a gg_boost_vimp object
autoplot(<gg_boost_effect>) plot(<gg_boost_effect>): Plot a gg_boost_effect object
autoplot(<gg_boost_calibration>) plot(<gg_boost_calibration>): Plot a gg_boost_calibration object

Shortcut — straight from a fit

Skips the two steps for the one figure worth looking at first. Every other figure is reached through its own extractor.

autoplot(<boostmtree>): Plot a boostmtree fit

Package overview

The two-step shape – extract, then render – and why the seam is where it is.

ggBoostedTrees ggBoostedTrees-package: ggBoostedTrees: Visually Exploring Boosted Tree Models