def get_tfg():
"""Get a FOOOFGroup object, with some fit power spectra, for testing."""
n_spectra = 2
xs, ys = gen_group_power_spectra(n_spectra, *default_group_params())
tfg = FOOOFGroup()
tfg.fit(xs, ys)
return tfg
def test_fg_report(skip_if_no_mpl):
"""Check that running the top level model method runs."""
n_spectra = 2
xs, ys = gen_group_power_spectra(n_spectra, *default_group_params())
tfg = FOOOFGroup()
tfg.report(xs, ys)
assert tfg
def test_fit_fooof_group_3d(tfg):
n_spectra = 2
xs, ys, _ = gen_group_power_spectra(n_spectra, *default_group_params())
ys = np.stack([ys, ys], axis=0)
tfg = FOOOFGroup()
fgs = fit_fooof_group_3d(tfg, xs, ys)
assert len(fgs) == 2
for fg in fgs:
assert fg
def test_fg_fit_par():
"""Test FOOOFGroup fit, running in parallel."""
n_spectra = 2
xs, ys = gen_group_power_spectra(n_spectra, *default_group_params())
tfg = FOOOFGroup()
tfg.fit(xs, ys, n_jobs=2)
out = tfg.get_results()
assert out
assert len(out) == n_spectra
assert isinstance(out[0], FOOOFResult)
assert np.all(out[1].background_params)
def test_fg_fit():
"""Test FOOOFGroup fit, no knee."""
n_spectra = 2
xs, ys, _ = gen_group_power_spectra(n_spectra, *default_group_params())
tfg = FOOOFGroup()
tfg.fit(xs, ys)
out = tfg.get_results()
assert out
assert len(out) == n_spectra
assert isinstance(out[0], FOOOFResults)
assert np.all(out[1].aperiodic_params)
from fooof.analysis import get_band_peak, get_band_peak_group
###############################################################################
# Reload some data and fit a FOOOF model to use
freqs = np.load('dat/freqs_lfp.npy')
spectrum = np.load('dat/spectrum_lfp.npy')
fm = FOOOF(peak_width_limits=[2, 8])
fm.fit(freqs, spectrum, [3, 30])
###############################################################################
# Generate some synthetic power spectra and fit a FOOOFGroup to use
freqs, spectra, _ = gen_group_power_spectra(
n_spectra=10,
freq_range=[3, 40],
background_params=param_sampler([[20, 2], [35, 1.5]]),
gauss_params=param_sampler([[], [10, 0.5, 2]]))
fg = FOOOFGroup(peak_width_limits=[1, 8],
min_peak_amplitude=0.05,
max_n_peaks=6,
verbose=False)
fg.fit(freqs, spectra)
###############################################################################
# FOOOF Analysis Utilities
# ------------------------
#
# FOOOF is packaged with minimal analysis utility functions.
#
# The plan is for the FOOOF module to stay this way, as supporting further analysis of FOOOF-derived results is largely outside the scope of the current module.
# - If there is a clear knee, then use knee fits.
# - This is likely across larger fitting ranges such as 1-150 Hz.
# - Be wary of ambiguous ranges, where there may or may not be a knee.
# - Trying to fit without a knee, when there is not a single consistent aperiodic signal, can lead to very bad fits. But it is also a known issue that trying to fit with a knee can lead to suboptimal fits when no knee is present.
# - We therefore currently recommend picking frequency ranges in which the expected aperiodic signal process is relatively clear.
###############################################################################
# Checking Fits Across a Group
# ----------------------------
# Set the parameters options for aperiodic signal and Gaussian peaks
bgp_opts = param_sampler([[20, 2], [50, 2.5], [35, 1.5]])
gauss_opts = param_sampler([[], [10, 0.5, 2], [10, 0.5, 2, 20, 0.3, 4]])
# Generate a group of power spectra
freqs, power_spectra, syn_params = gen_group_power_spectra(10, [3, 50], bgp_opts, gauss_opts)
###############################################################################
# Initialize a FOOOFGroup
fg = FOOOFGroup(peak_width_limits=[1, 6])
###############################################################################
# Fit FOOOF and report on the group
fg.report(freqs, power_spectra)
###############################################################################
# Find the index of the worst FOOOF fit from the group
worst_fit_ind = np.argmax(fg.get_all_data('error'))
# Set some options for peak parameters # Generated power spectra will have either no peaks, a 10 Hz peak, or a 10 Hz & 20 Hz peak gauss_opts = param_sampler([[], [10, 0.5, 2], [10, 0.5, 2, 20, 0.3, 4]]) ############################################################################### # # We can now feed these settings into :func:`gen_group_power_spectra`, that will generate a group of power spectra for us. # # Note that this function also returns a list of the parameters used to generate each power spectrum. ############################################################################### # Generate the group of synthetic spectra # Note that this function also returns a list of the parameters for each func freqs, spectra, syn_params = gen_group_power_spectra(n_psds, f_range, bgp_opts, gauss_opts) ############################################################################### # FOOOFGroup # ---------- # # The FOOOFGroup object is very similar to the FOOOF object (programmatically, it inherits from the FOOOF object), and can be used in the same way. # # The main difference is that instead of running across a single power spectrum, it operates across 2D matrices containing multiple power spectra. # # Note that by 'group' we mean merely to refer to a group of power-spectra, not necessarily to a group in terms of multiple subjects or similar. Most likely, a FOOOFGroup will be run across a collection of spectra from across channels, and/or across trials, within or across subjects. # # The main difference with the FOOOFGroup object, is that it also contains a 'power_spectra' attribute, which stores the matrix of power-spectra to be fit, and collects fit results into a 'group_results' attribute. Otherwise, FOOOFGroup supports all the same functionality, accessed in the same way as the FOOOF object. # # Internally, it runs the exact same fitting procedure, per spectrum, as the FOOOF object.
