Calculate ALE data — calc

This function is not exported. It is a complete reimplementation of the ALE algorithm relative to the reference in ALEPlot::ALEPlot(). In addition to adding bootstrapping and handling of categorical y variables, it reimplements categorical x interactions.

Usage

calc_ale(
  data,
  model,
  x_cols,
  y_col,
  y_cats,
  pred_fun,
  pred_type,
  max_num_bins,
  boot_it,
  seed,
  boot_alpha,
  boot_centre,
  boot_ale_y = FALSE,
  bins = NULL,
  ns = NULL,
  ale_y_norm_funs = NULL,
  p_dist = NULL
)

Arguments

data: See documentation for ale()
model: See documentation for ale()
x_cols: character(1 or 2). Names of columns in X for which ALE data is to be calculated. Length 1 for 1D ALE and length 2 for 2D ALE.
y_col: character(1). Name of the target y column.
y_cats: character. The categories of y. For most cases with non-categorical y, y_cats == y_col.
pred_fun: See documentation for ale()
pred_type: See documentation for ale()
max_num_bins: See documentation for ale()
boot_it: See documentation for ale()
seed: See documentation for ale()
boot_alpha: See documentation for ale()
boot_centre: See documentation for ale()
boot_ale_y: logical(1). If TRUE, return the bootstrap matrix of ALE y values. If FALSE (default) return NULL for the boot_ale_y element of the return value.
bins, ns: numeric or ordinal vector,integer vector. Normally generated automatically (if bins == NULL), but if provided, the provided values will be used instead. They would mainly be provided from model_bootstrap().
ale_y_norm_funs: list of functions. Custom functions for normalizing ALE y for statistics. It is usually a list(1), but for categorical y, there is a distinct function for each y category. If provided, ale_y_norm_funs saves some time since it is usually the same for all all variables throughout one call to ale(). For now, used as a flag to determine whether statistics will be calculated or not; if NULL, statistics will not be calculated.
p_dist: See documentation for p_values in ale()

Details

For details about arguments not documented here, see ale().

References

Apley, Daniel W., and Jingyu Zhu. "Visualizing the effects of predictor variables in black box supervised learning models." Journal of the Royal Statistical Society Series B: Statistical Methodology 82.4 (2020): 1059-1086.

Okoli, Chitu. 2023. “Statistical Inference Using Machine Learning and Classical Techniques Based on Accumulated Local Effects (ALE).” arXiv. doi:10.48550/arXiv.2310.09877.