bradley_terry_tree

heterogeneity.bradley_terry_tree(matrix, method_names, dataset_names, numeric_features=None, categorical_features=None, polarity='higher_is_better', minsize=5, alpha=0.05)

Fit a Bradley-Terry tree on per-dataset paired method comparisons.

Parameters

Name	Type	Description	Default
matrix	np.ndarray	Array of shape (n_methods, n_datasets) holding one metric’s scores.	required
method_names	Sequence[str]	Length n_methods row labels, the objects compared.	required
dataset_names	Sequence[str]	Length n_datasets column labels, the subjects whose features split the tree.	required
numeric_features	dict[str, Sequence[float]] | None	Maps from feature name to a length n_datasets sequence of values, the candidate splitting variables. At least one feature across the two maps is required.	None
categorical_features	dict[str, Sequence[float]] | None	Maps from feature name to a length n_datasets sequence of values, the candidate splitting variables. At least one feature across the two maps is required.	None
polarity	str	"higher_is_better" or "lower_is_better", the metric direction used to orient the pairwise comparisons.	'higher_is_better'
minsize	int	Minimal number of datasets in a node; the tree will not create a leaf smaller than this.	5
alpha	float	Significance level for the parameter-stability split test.	0.05

Returns

Type	Description
BradleyTerryTreeReport

Raises

Type	Description
ValueError	For shape, length, polarity, or empty-feature problems.
RNotAvailableError	If the R toolchain with psychotree is not available.
RExecutionError	If the R subprocess fails.