average or sum, weighted or unweighted, over axes#

Summary#

Compute a weighted or unweighted mean/sum over the given axes, propagate existing uncertainties, and optionally add scatter-derived uncertainty estimates.

Metadata#

Import path: modacor.modules.base_modules.reduce_dimensionality.ReduceDimensionality
Source: src/modacor/modules/base_modules/reduce_dimensionality.py
Module ID: ReduceDimensionality
Module version: 20260927.1
Keywords: average, mean, weighted, nanmean, reduce, axis, sum

Required data keys#

signal

Modifies#

signal: signal, uncertainties, units, weights

Required arguments#

None

Default configuration#

{
  "axes": null,
  "mask_bits": null,
  "mask_key": null,
  "nan_policy": "omit",
  "reduction": "mean",
  "uncertainty_estimation": null,
  "use_weights": true
}

Argument specification#

Argument	Type	Required	Default	Dependency role	Description
`axes`	int or list or tuple or str or NoneType	No	-	-	Axis or axes to reduce (int, list/tuple, or None for all). Use ‘non_data’ to reduce every leading axis before the final rank_of_data axes.
`mask_bits`	int or list or tuple or NoneType	No	-	-	Optional uint32 bit value or iterable of bit values to exclude. None excludes any nonzero mask.
`mask_key`	str or NoneType	No	-	-	Optional BaseData mask key in each selected DataBundle. Nonzero mask values are excluded without modifying the input signal.
`nan_policy`	str	No	omit	-	NaN handling policy: ‘omit’ or ‘propagate’.
`reduction`	str	No	mean	-	Reduction method: ‘mean’ or ‘sum’.
`uncertainty_estimation`	dict or NoneType	No	-	-	Optional mapping containing estimator definitions and a collision policy for scatter-derived uncertainty components.
`use_weights`	bool	No	True	-	Use BaseData weights for weighted reduction.

References#

DOI 10.1088/0953-8984/25/38/383201

Notes#

This step reduces the dimensionality of the signal by averaging over one or more axes. With axes=’non_data’, it automatically reduces leading acquisition axes until signal.ndim equals rank_of_data. An optional integer mask can exclude selected reason bits without modifying the input signal. Units are preserved; complete axes metadata is reduced along the same axes.

Table of Contents