ParameterCorrelation

class nyx.infer.uncertainty.ParameterCorrelation

Bases: object

How a fit’s parameters trade off against one another.

names has one entry per free scalar in the optimizer’s flattening order, so array parameters repeat, and sigma is in that same stepping space. n_unconstrained counts the directions the pseudoinverse discarded: non-zero means the data does not determine some combination.

Attributes Summary

correlation
covariance
n_unconstrained
names
sigma

Methods Summary

worst_pairs([count, threshold])

The count most correlated pairs of distinct parameters.

Attributes Documentation

correlation = <dataclasses._MISSING_TYPE object>

covariance = <dataclasses._MISSING_TYPE object>

n_unconstrained = <dataclasses._MISSING_TYPE object>

names = <dataclasses._MISSING_TYPE object>

sigma = <dataclasses._MISSING_TYPE object>

Methods Documentation

worst_pairs(count=10, *, threshold=0.0)

The count most correlated pairs of distinct parameters.

Reported per parameter, not per scalar: an array parameter contributes only the strongest correlation in its block.

Parameters:

• count (int)

• threshold (float) – Drop pairs correlated more weakly than this.

Return type:

list of tuple

__init__(names, covariance, correlation, sigma, n_unconstrained)

classmethod __new__(*args, **kwargs)