Statistics Functions#

EEGPrep statistics helpers mirror the useful parts of EEGLAB’s functions/statistics package while using explicit NumPy contracts. Unless noted otherwise, condition arrays use their final axis for cases, subjects, or surrogate replications.

Condition Tests#

eegprep.functions.statistics.statcond(data, *, paired='auto', method='param', mode=None, naccu=200, variance='homogenous', forceanova=False, tail='both', axis=-1, rng=None, alpha=None, surrog=None, stats=None, return_resampling_array=False)

Compare condition arrays using EEGLAB-style t-tests or ANOVAs.

Parameters:

  • data (Any) – One- or two-dimensional sequence of condition arrays. The case dimension is the last axis by default.

  • paired (str | bool) – "auto", "on"/True, or "off"/False.

  • method (str) – "param", "perm", or "bootstrap".

  • mode (str | None) – Legacy alias for method.

  • naccu (int) – Number of surrogate samples for nonparametric methods.

  • variance (str) – "homogenous" or "inhomogenous" for unpaired t-tests.

  • forceanova (bool) – Use one-way ANOVA instead of a two-condition t-test.

  • tail (str) – Empirical-tail mode for supplied or computed surrogates.

  • axis (int) – Axis in each condition array that stores cases.

  • rng (Generator | int | None) – Optional NumPy generator or seed for deterministic resampling.

  • alpha (float | None) – Optional threshold for confidence intervals and masks; requires a nonparametric method or supplied surrogate statistics.

  • surrog (Any) – Precomputed surrogate statistic array.

  • stats (Any) – Observed statistic to pair with surrog.

  • return_resampling_array (bool) – Return surrogate condition grids instead of computing statistics.

Return type:

StatcondResult | SurrogateDistribution

eegprep.functions.statistics.ttest_cell(a, b=None, *, axis=-1)

Compute paired t-statistics across the case axis.

Parameters:
Return type:

tuple[ndarray, int]

eegprep.functions.statistics.ttest2_cell(a, b=None, variance='homogenous', *, axis=-1)

Compute unpaired t-statistics across the case axis.

Parameters:
Return type:

tuple[ndarray, ndarray | int]

eegprep.functions.statistics.anova1_cell(data, *, axis=-1)

Compute one-way unpaired ANOVA F-statistics across condition arrays.

Parameters:
Return type:

tuple[ndarray, tuple[int, int]]

eegprep.functions.statistics.anova1rm_cell(data, *, axis=-1)

Compute one-way repeated-measures ANOVA F-statistics.

Parameters:
Return type:

tuple[ndarray, tuple[int, int]]

eegprep.functions.statistics.anova2_cell(data, *, axis=-1)

Compute balanced two-way unpaired ANOVA F-statistics.

Parameters:
Return type:

TwoWayAnovaResult

eegprep.functions.statistics.anova2rm_cell(data, *, axis=-1)

Compute two-way repeated-measures ANOVA F-statistics.

Parameters:
Return type:

TwoWayAnovaResult

Multiple Comparisons And Surrogates#

eegprep.functions.statistics.fdr(pvals, q=None, fdr_type='parametric')

Compute Benjamini-Hochberg or Benjamini-Yekutieli FDR thresholds.

Parameters:

  • pvals (Any) – Numeric p-value array with values in the closed interval [0, 1].

  • q (float | None) – Desired false discovery rate. If omitted, an EEGLAB-style array of corrected thresholds is returned.

  • fdr_type (str) – "parametric" for Benjamini-Hochberg or "nonparametric"/"nonParametric" for Benjamini-Yekutieli.

Returns:

Threshold and boolean mask with the same shape as pvals.

Return type:

FDRResult

eegprep.functions.statistics.stat_surrogate_pvals(distribution, observed, tail='both')

Compute empirical p-values against a surrogate distribution.

Parameters:

  • distribution (Any) – Surrogate statistic array whose last axis stores surrogate replications.

  • observed (Any) – Observed statistic array matching distribution.shape[:-1].

  • tail (str) – "right"/"upper"/"one", "left"/"lower", or "both".

Return type:

ndarray

eegprep.functions.statistics.stat_surrogate_ci(distribution, alpha=0.05, tail='both')

Compute surrogate confidence intervals along the last axis.

Parameters:

  • distribution (Any) – Surrogate statistic array whose last axis stores surrogate replications.

  • alpha (float) – Type-I error rate.

  • tail (str) – "upper", "lower", "one", or "both".

Return type:

ndarray

eegprep.functions.statistics.surrogdistrib(data, *, method='perm', pairing='on', naccu=1, axis=-1, rng=None)

Build bootstrap or permutation surrogate condition grids.

Parameters:

  • data (Any) – One- or two-dimensional sequence of condition arrays.

  • method (str) – "perm"/"permutation" or "bootstrap".

  • pairing (str) – "on" to preserve case identity across conditions or "off" to resample from the pooled case axis.

  • naccu (int) – Number of surrogate grids to generate.

  • axis (int) – Axis in each condition array that stores cases.

  • rng (Generator | int | None) – Optional NumPy generator or seed for deterministic resampling.

Return type:

SurrogateDistribution

Data Helpers#

eegprep.functions.statistics.concatdata(data, *, axis=-1)

Concatenate condition arrays along their case axis.

Parameters:

  • data (Any) – One- or two-dimensional sequence of condition arrays.

  • axis (int) – Axis in each condition array that stores cases.

Return type:

ConcatenatedData

eegprep.functions.statistics.corrcoef_cell(a, b=None, *, axis=-1)

Compute pairwise correlations along a case axis.

Parameters:
Return type:

ndarray

eegprep.functions.statistics.teststat(seed=0)

Run deterministic smoke checks for the EEGPrep statistics package.

Parameters:

seed (int)

Return type:

dict[str, float]