[FIX] collateral effects when scaling of data in _compute_covariance_…

…auto (mne-tools#5698)

doc/whats_new.rst

@@ -107,6 +107,9 @@ Changelog
 
 Bug
 ~~~
+
+- Fix: bug with scaling of data in func:`mne.cov._compute_covariance_auto` by `David Sabbagh `_ 
+
 - Fix :func:`mne.io.Raw.plot_projs_topomap` by `Joan Massich`_
 
 - Fix bug in :func:`mne.minimum_norm.compute_source_psd` where the ``stc.times`` output was scaled by 1000, by `Eric Larson`_
@@ -2980,3 +2983,5 @@ of commits):
 .. _Rasmus Zetter: https://people.aalto.fi/rasmus.zetter
 
 .. _Marcin Koculak: https://github.com/mkoculak
+
+.. _David Sabbagh: https://github.com/DavidSabbagh

examples/decoding/plot_decoding_spoc_CMC.py

@@ -36,7 +36,7 @@
 from sklearn.linear_model import Ridge
 from sklearn.model_selection import KFold, cross_val_predict
 
-# define parameters
+# Define parameters
 fname = data_path() + '/SubjectCMC.ds'
 raw = mne.io.read_raw_ctf(fname)
 raw.crop(50., 250.).load_data()  # crop for memory purposes
@@ -45,7 +45,7 @@
 emg = raw.copy().pick_channels(['EMGlft'])
 emg.filter(20., None, fir_design='firwin')
 
-# Filter MEG data to focus on alpha band
+# Filter MEG data to focus on beta band
 raw.pick_types(meg=True, ref_meg=True, eeg=False, eog=False)
 raw.filter(15., 30., fir_design='firwin')
 
@@ -70,7 +70,7 @@
 # Run cross validaton
 y_preds = cross_val_predict(clf, X, y, cv=cv)
 
-# plot the True EMG power and the EMG power predicted from MEG data
+# Plot the True EMG power and the EMG power predicted from MEG data
 fig, ax = plt.subplots(1, 1, figsize=[10, 4])
 times = raw.times[meg_epochs.events[:, 0] - raw.first_samp]
 ax.plot(times, y_preds, color='b', label='Predicted EMG')

mne/cov.py

@@ -989,6 +989,7 @@ def _compute_covariance_auto(data, method, info, method_params, cv,
         logliks = [None]
 
     # undo scaling
+    _undo_scaling_array(data.T, picks_list=picks_list, scalings=scalings)
     for c in estimator_cov_info:
         _undo_scaling_cov(c[1], picks_list, scalings)
 
@@ -997,8 +998,9 @@ def _compute_covariance_auto(data, method, info, method_params, cv,
     cov_methods = [c.__name__ if callable(c) else c for c in method]
     runtime_infos, covs = list(runtime_infos), list(covs)
     my_zip = zip(cov_methods, runtime_infos, logliks, covs, estimators)
-    for this_method, runtime_info, loglik, data, est in my_zip:
-        out[this_method] = {'loglik': loglik, 'data': data, 'estimator': est}
+    for this_method, runtime_info, loglik, cov_data, est in my_zip:
+        out[this_method] = {
+            'loglik': loglik, 'data': cov_data, 'estimator': est}
         if runtime_info is not None:
             out[this_method].update(runtime_info)
 

mne/tests/test_cov.py

@@ -15,7 +15,8 @@
 from mne.cov import (regularize, whiten_evoked, _estimate_rank_meeg_cov,
                      _auto_low_rank_model, _apply_scaling_cov,
                      _undo_scaling_cov, prepare_noise_cov, compute_whitener,
-                     _apply_scaling_array, _undo_scaling_array)
+                     _apply_scaling_array, _undo_scaling_array,
+                     _regularized_covariance)
 
 from mne import (read_cov, write_cov, Epochs, merge_events,
                  find_events, compute_raw_covariance,
@@ -496,6 +497,10 @@ def test_cov_scaling():
     _undo_scaling_array(data, picks_list, scalings=scalings)
     assert_allclose(data, evoked.data, atol=1e-20)
 
+    # check that input data remain unchanged. gh-5698
+    _regularized_covariance(data)
+    assert_array_almost_equal(data, evoked.data)
+
 
 @requires_version('sklearn', '0.15')
 def test_auto_low_rank():