Skip to contents

Spectral parameterization (specparam / FOOOF)

Source: R/ops-specparam.R
specparam.Rd

Parameterizes each channel's power spectrum into an aperiodic 1/f component (offset, exponent, and an optional knee) plus a set of Gaussian periodic peaks, following the specparam / FOOOF algorithm (Donoghue et al. 2020). The aperiodic exponent is a marker of excitation/inhibition balance and cortical state; the peaks capture genuine oscillations separated from the 1/f background. The fit runs on the Welch PSD computed by the same machinery as bandPower().

Usage

specparam(
  x,
  freq_range = c(1, 45),
  peak_width_limits = c(1, 12),
  max_n_peaks = 6L,
  aperiodic_mode = c("fixed", "knee"),
  min_peak_height = 0.05,
  peak_threshold = 2,
  nperseg = NULL,
  assay_name = NULL
)

# S3 method for class 'specparam_result'
print(x, ...)

Arguments

x

A specparam_result.

freq_range

Numeric length-2 frequency range in Hz to fit (default: c(1, 45)).

peak_width_limits

Numeric length-2 minimum and maximum peak bandwidth in Hz (default: c(1, 12)).

max_n_peaks

Maximum number of periodic peaks to fit (default: 6).

aperiodic_mode

"fixed" for a pure 1/f (offset, exponent) or "knee" to also fit a knee frequency.

min_peak_height

Minimum peak height above the aperiodic fit, in log10 power (default: 0.05).

peak_threshold

Minimum peak height in units of the standard deviation of the flattened spectrum (default: 2).

nperseg

Welch segment length in samples (default: about a 2-second segment for adequate low-frequency resolution).

assay_name

Input assay name (default: the default assay).

...

Ignored.

Value

An object of class "specparam_result": a list with aperiodic (a data.frame of per-channel offset, exponent, and knee), peaks (a data.frame of per-channel center frequency CF, power PW, and bandwidth BW), fit (per-channel R-squared and error), spectra (per-channel log spectra and fitted model for plotting), and the settings used.

References

Donoghue, T., et al. (2020). Parameterizing neural power spectra into periodic and aperiodic components. Nature Neuroscience, 23(12), 1655-1665.

See also

plotSpecparam(), specparamBiomarker(), bandPower()

Examples

if (FALSE) { # \dontrun{
pe <- make_pe_2d(n_time = 4000, n_channels = 4, sr = 250)
sp <- specparam(pe, freq_range = c(1, 40))
sp$aperiodic
} # }