parameter_spread

SuperLSS.parameter_spread(

X: object | EagerFrame,

*,

threshold: float,

sample_weight: NDArray | None = None,

offsets: Mapping[str, NDArray] | None = None,

**kwargs: Any,

) → Spread

Return how far the fitted parameters spread, and how far identical prices do.

The histograms show the sharpness of each fitted parameter and of the predicted tail quantile over the rows; the identically-priced table bins rows by predicted mean and reports the spread of P(Y > threshold) inside each bin, which is the quantity a location-only model has no way to distinguish. Under prior semantics sample_weight enters twice, because it means the same thing in both places: it weighs the ratio of sums the table reports and it is part of each row’s own law. Frequency weights replicate bins, percentiles and histogram counts. Zero-weight rows and their offsets are omitted before prediction.