Grouped spectrum with resampled null testing — group_spectrum

Builds a null distribution for the group-average spectrum by resampling each subject series, then computes empirical p-values, z-scores, and optional FDR correction across frequencies.

Usage

group_spectrum_test(
  x,
  fs,
  flim = NULL,
  null = c("shuffle_labels", "circular_shift"),
  nrep = 1000,
  detrend = TRUE,
  detrend_order = 1L,
  pad_to = NULL,
  spectrum = c("amplitude", "raw_power"),
  taper = c("none", "hann", "hanning"),
  p_adjust = c("none", "fdr"),
  fcor = FALSE,
  seed = NULL
)

Arguments

x: numeric matrix with shape subjects x bins, or a grouped-series object.
fs: sampling rate in Hz along the bin axis.
flim: optional length-2 frequency bounds.
null: null generator; one of "shuffle_labels" or "circular_shift".
nrep: number of null draws.
detrend: logical; if TRUE, remove a linear trend from each subject series before spectral estimation.
detrend_order: non-negative integer polynomial order used when detrend = TRUE.
pad_to: optional FFT length used for zero-padding each subject series before spectral estimation.
spectrum: spectrum scaling passed to spectral_peak.
taper: taper passed to spectral_peak.
p_adjust: p-value adjustment; one of "none" or "fdr".
fcor: logical; apply 1/f correction to each subject spectrum.
seed: optional random seed for reproducible null draws.

Value

list with observed, null, stats, and meta.