spectral_connectivity.statistics.JackknifeResult

class JackknifeResult(estimate: ndarray[tuple[int, ...], dtype[floating]], bias_corrected: ndarray[tuple[int, ...], dtype[floating]], standard_error: ndarray[tuple[int, ...], dtype[floating]], confidence_interval: tuple[ndarray[tuple[int, ...], dtype[floating]], ndarray[tuple[int, ...], dtype[floating]]], n_observations: int, transformation: str)

Bases: object

Leave-one-observation-out estimate and Student-t jackknife interval.

Every array attribute has the shape of the underlying measure, e.g. (n_time, n_nonnegative_frequencies, n_signals, n_signals) for a pairwise connectivity measure.

estimate

The full-sample estimate on the original scale.

Type:

array

bias_corrected

Jackknife bias-corrected estimate on the original scale.

Type:

array

standard_error

Standard error on the original scale (delta method through the transformation).

Type:

array

confidence_interval

Confidence bounds on the original scale, formed on the transformed scale as estimate -/+ t * standard_error with the Student t critical value on n_observations - 1 degrees of freedom. For the "circular" transformation the bounds are wrapped to (-pi, pi]: lower > upper means the interval crosses +/-pi and is [lower, pi] U (-pi, upper], and a bin whose half-width reaches pi (the whole circle, phase unresolved) is reported as (-pi, pi).

Type:

tuple of (lower, upper) arrays

n_observations

Number of leave-one-out replicates; the interval’s critical value has n_observations - 1 degrees of freedom.

Type:

int

transformation

Variance-stabilizing transformation used for the interval.

Type:

str

Methods

Attributes

estimate
bias_corrected
standard_error
confidence_interval
n_observations
transformation