spacr.regression_families

Describe and group the statistical assumptions of regression families.

The available families are organized into three methodological groups:

parametric

the response is given a distribution and the p-values come from that distribution’s theory. Least squares, the GLM links, the penalised fits and the mixed model.

robust_semiparametric

linear models whose coefficients remain parametric but whose loss reduces sensitivity to extreme wells. rlm, huber and quantile belong to this group; quantile regression leaves the error distribution unspecified.

rank_based

fits based on well order rather than response magnitudes. rra is the nonparametric family in the current inventory.

Distribution-free inference is independent of these effect-estimation groups. Any coefficient-producing family can be paired with a permutation null. The stored regression-type values are retained for settings-file compatibility.

Functions

family_group(→ str)	Which of the three kinds a family is.
family_label(→ str)	The one line a dropdown shows for a family: its kind and its assumption.
regression_family_choices(→ List[Tuple[str, str]])	Every fittable family as (stored value, label), grouped.

Module Contents

spacr.regression_families.family_group(name: str) → str

Which of the three kinds a family is.

Parameters:

name – a stored regression_type value.

Returns:

'parametric', 'robust_semiparametric' or 'rank_based'.

Raises:

KeyError – a family this table does not place, which is the signal that the inventory grew and this did not.

spacr.regression_families.family_label(name: str) → str

The one line a dropdown shows for a family: its kind and its assumption.

Parameters:

name – stored regression-family value to label.

Returns:

the stored value, family-group title, and statistical assumption as one menu label.

Raises:

KeyError – if name has no registered assumption or family group.

The stored value leads, because that is what a user is looking for and what every settings file and results folder is named after.

spacr.regression_families.regression_family_choices() → List[Tuple[str, str]]

Every fittable family as (stored value, label), grouped.

Parametric first, then robust/semiparametric, then rank-based; the stored value is unchanged, so a panel that renders these writes exactly what it wrote before and every settings file already on disk still means what it meant.

Returns:

ordered (stored value, label) pairs for every fittable regression family.