CrossValidationResult#

class superglm.CrossValidationResult( fold_scores: DataFrame, mean_scores: dict[str, float], pooled_scores: dict[str, float], std_scores: dict[str, float], fold_indices: list[tuple[NDArray, NDArray]] | None = None, curve_similarity: dict[str, Any] | None = None, oof_predictions: NDArray | None = None, estimators: list | None = None, )#

Bases: object

Structured result from superglm.cross_validate().

Attributes:

fold_scoresDataFrame: One row per fold with columns: fold, n_train, n_test, fit_time_s, score_time_s, converged, n_iter, effective_df, plus one column per requested metric.
mean_scoresdict: Equal-weight mean of each per-fold metric across folds. Built-in deviance and negative log-likelihood are normalized within each fold by the likelihood size the model’s declared weight_semantics implies: sum(sample_weight) under "frequency", the count of positive-weight validation rows under "prior".
pooled_scoresdict: Supported overall pooled metrics, computed as ratio-of-sums rather than mean-of-fold-ratios, with the same denominator.
std_scoresdict: Standard deviation of each metric across folds.
fold_indiceslist[tuple[ndarray, ndarray]] or None: Per-fold (train_idx, test_idx) pairs from the CV splitter.
curve_similaritydict or None: Fold-by-fold term similarity diagnostics for comparable main effects.
oof_predictionsndarray or None: Out-of-fold predictions (response scale), same length as y. None unless return_oof=True.
estimatorslist or None: Fitted model per fold. None unless return_estimators=True.

plot_terms_by_fold(

X: object,

*,

sample_weight: NDArray | None = None,

terms: str | list[str] | None = None,

engine: str = 'plotly',

**kwargs,

)#: Plot fold-specific main effects using the shared comparison engine.