Plot SHAP values vs feature values — plot_shap

Plot exact XGBoost SHAP values against feature values across all training samples for one perturbation. Optionally highlight selected training samples or user samples predicted with add_powerup_predictions().

Usage

plot_shap_scatter(
  models,
  prepared,
  perturbation,
  n_features = 6,
  features = NULL,
  samples = NULL,
  source = "training",
  n_columns = 3,
  sample_colors = NULL
)

Arguments

models: A named list of models returned by fit_powerup_models().
prepared: The prepared POWERUP object used for model training and prediction.
perturbation: A single perturbation name to plot.
n_features: Number of top contributing features to plot when features is NULL.
features: Optional character vector of specific model features to plot.
samples: Optional character vector of samples to highlight.
source: Either "training" (default) or "user". User samples require models that have been passed through add_powerup_predictions().
n_columns: Number of columns in the resulting plot grid.
sample_colors: Optional colors for highlighted samples.

Value

A combined ggplot object containing SHAP dependence panels for the selected features.

Examples

if (FALSE) { # \dontrun{
plot_shap_scatter(
  models,
  prepared,
  perturbation = "CTNNB1"
)

plot_shap_scatter(
  models,
  prepared,
  perturbation = "CTNNB1",
  samples = "ACH-000957",
  source = "user"
)
} # }