neurotools.stats.covalign module
Covariance alignment routines related to
- neurotools.stats.covalign.keeprank(Σ, eps=0.1)[source]
Return all eigenvectors with eigenvalues greater than epsilon times the largest eigenvalue.
- Parameters:
Σ (2D square array-like (positive semidefinite matrix)) – Covariance matrix
- Return type:
Leading eigenvectors of Σ.
- neurotools.stats.covalign.expected_intersection_rank1(Δμ, Σ, N=100)[source]
Rank-1 approximation of average overlap. Uses only the largest eigenvector.
- Parameters:
Δμ (1D array-like (vector)) – Direction of shift in mean activity
Σ (2D square array-like (positive semidefinite matrix)) – Covariance matrix of distribution of single-sessiona activity
- neurotools.stats.covalign.expected_intersection(Δμ, Σ)[source]
Unnormalized expected intersection.
- Parameters:
Δμ (1D array-like (vector)) – Direction of shift in mean activity
Σ (2D square array-like (positive semidefinite matrix)) – Covariance matrix of distribution of single-sessiona activity
- neurotools.stats.covalign.expected_intersection_enormed(Δμ, Σ)[source]
Normalized expected intersection.
- Parameters:
Δμ (1D array-like (vector)) – Direction of shift in mean activity
Σ (2D square array-like (positive semidefinite matrix)) – Covariance matrix of distribution of single-sessiona activity
- Returns:
Root-mean-square normalized magnitude of the dot-product between the
vector Δμ and unit vector directions of random variables sampled from
a Gaussian with covariance Σ.
- neurotools.stats.covalign.expected_intersection_enormed_chance(Δμ, Σ, N=300, pct=95)[source]
Normalized expected intersection chance level.
- Parameters:
Δμ (1D array-like (vector)) – Direction of shift in mean activity
Σ (2D square array-like (positive semidefinite matrix)) – Covariance matrix of distribution of single-sessiona activity
N (int, defaults to 300) – The number of random Monte-Carlo samples to use to asses chance level.
pct (numeric in (0,100), defaults to 90) – The chance-level percentile to report
- Returns:
The pct percentile chance level, via Monte-Carlo sampling, for the
null hypothesis that Δμ has no more than chance alignment with the
structure of Σ.
- neurotools.stats.covalign.rebuild_unit_quality_caches()[source]
Rebuilds a disk cache of which units are “good” units (have viable recordings on a given session).
- neurotools.stats.covalign.get_orthogonal_alignment(animal, CUE, PREV, verbose=True, **kwargs)[source]
Compute orthogonal complement drift alignment statistics for the given animal, previous cue, and current cue.
- neurotools.stats.covalign.alignment_angle(Δμ, Σ)[source]
Computed generalzation of alignment angle between a vector and a covariance matrix
- neurotools.stats.covalign.sample_alignment_angle(Σ, K=10)[source]
Computed generalzation of alignment angle between a vector and a covariance matrix
- neurotools.stats.covalign.sample_alignment_self(Σ, K=10)[source]
Self-alignment of distribution Sanity check for baseline
- neurotools.stats.covalign.sample_alignment_cross(Σ, Σother, K=10)[source]
Self-alignment of distribution Sanity check for baseline