Averaging ERPs: Creating MNE Evoked objects

Averaging ERPs: Creating MNE Evoked objects#

Since averaging across trials is typically the end goal of an ERP experiment, MNE has a distinct class, Evoked, for ERP data where multiple trials have been averaged for each experimental condition or trial type.

We’ll firstimport MNE and the mastoid-referenced, cleaned epochs we saved in the previous step.

import mne
mne.set_log_level('error')  # reduce extraneous MNE output

# Participant ID code
p_id = 'N400_dataset_14'
data_dir = 'data'

epochs_mastoidref = mne.read_epochs(data_dir + '/' + p_id + '-mstd-epo.fif', preload=True)

Create Evoked objects#

Unlike Epochs, which by definition contain multiple trials and in which each trial has an associated condition label (event code), Evoked data sets are averages across trials for a single condition. Thus we have to create a separate Evoked object for each condition in our experiment. For convenience, we can store these in a list, and we can created the list of Evoked objects using dictionary comprehension, so that each dictionary key is the condition name and its value is the Evoked object:

conditions = ['Match', 'Mismatch']

evokeds = {c:epochs_mastoidref[c].average() for c in conditions}

evokeds    

{'Match': <Evoked | '0.25 × Match/A + 0.25 × Match/B + 0.25 × Match/C + 0.25 × Match/D' (average, N=72), -0.2 – 1 s, baseline -0.2 – 0 s, 16 ch, ~105 kB>,
 'Mismatch': <Evoked | '0.25 × Mismatch/A + 0.25 × Mismatch/B + 0.25 × Mismatch/C + 0.25 × Mismatch/D' (average, N=72), -0.2 – 1 s, baseline -0.2 – 0 s, 16 ch, ~105 kB>}

Plot average ERP for each condition#

We can use a method called plot_joint() to generate a combination of “butterfly” plots (waveforms of all electrodes overlaid) and topographic maps at specific time points that we specify. This is a nice way of relating scalp topographies and waveforms. The time points specified for the topo plots were selected based on early peaks apparent in the waveforms above, and then on 200 ms intervals over the time period when the N400 is expected:

times = [.150, .250, .400, .600, .800]
for c in evokeds.keys():
    evokeds[c].plot_joint(times=times, title=c);

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/IPython/core/formatters.py:340, in BaseFormatter.__call__(self, obj)
    338     pass
    339 else:
--> 340     return printer(obj)
    341 # Finally look for special method names
    342 method = get_real_method(obj, self.print_method)

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/IPython/core/pylabtools.py:152, in print_figure(fig, fmt, bbox_inches, base64, **kwargs)
    149     from matplotlib.backend_bases import FigureCanvasBase
    150     FigureCanvasBase(fig)
--> 152 fig.canvas.print_figure(bytes_io, **kw)
    153 data = bytes_io.getvalue()
    154 if fmt == 'svg':

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/matplotlib/backend_bases.py:2353, in FigureCanvasBase.print_figure(self, filename, dpi, facecolor, edgecolor, orientation, format, bbox_inches, pad_inches, bbox_extra_artists, backend, **kwargs)
   2350         bbox_inches = bbox_inches.padded(pad_inches)
   2352     # call adjust_bbox to save only the given area
-> 2353     restore_bbox = _tight_bbox.adjust_bbox(
   2354         self.figure, bbox_inches, self.figure.canvas.fixed_dpi)
   2356     _bbox_inches_restore = (bbox_inches, restore_bbox)
   2357 else:

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/matplotlib/_tight_bbox.py:28, in adjust_bbox(fig, bbox_inches, fixed_dpi)
     26 locator = ax.get_axes_locator()
     27 if locator is not None:
---> 28     ax.apply_aspect(locator(ax, None))
     29 locator_list.append(locator)
     30 current_pos = ax.get_position(original=False).frozen()

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/mpl_toolkits/axes_grid1/inset_locator.py:73, in AnchoredLocatorBase.__call__(self, ax, renderer)
     71 def __call__(self, ax, renderer):
     72     self.axes = ax
---> 73     bbox = self.get_window_extent(renderer)
     74     px, py = self.get_offset(bbox.width, bbox.height, 0, 0, renderer)
     75     bbox_canvas = Bbox.from_bounds(px, py, bbox.width, bbox.height)

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/matplotlib/offsetbox.py:399, in OffsetBox.get_window_extent(self, renderer)
    396 def get_window_extent(self, renderer=None):
    397     # docstring inherited
    398     if renderer is None:
--> 399         renderer = self.figure._get_renderer()
    400     bbox = self.get_bbox(renderer)
    401     try:  # Some subclasses redefine get_offset to take no args.

AttributeError: 'NoneType' object has no attribute '_get_renderer'

<Figure size 800x420 with 9 Axes>

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/IPython/core/formatters.py:340, in BaseFormatter.__call__(self, obj)
    338     pass
    339 else:
--> 340     return printer(obj)
    341 # Finally look for special method names
    342 method = get_real_method(obj, self.print_method)

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/IPython/core/pylabtools.py:152, in print_figure(fig, fmt, bbox_inches, base64, **kwargs)
    149     from matplotlib.backend_bases import FigureCanvasBase
    150     FigureCanvasBase(fig)
--> 152 fig.canvas.print_figure(bytes_io, **kw)
    153 data = bytes_io.getvalue()
    154 if fmt == 'svg':

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/matplotlib/backend_bases.py:2353, in FigureCanvasBase.print_figure(self, filename, dpi, facecolor, edgecolor, orientation, format, bbox_inches, pad_inches, bbox_extra_artists, backend, **kwargs)
   2350         bbox_inches = bbox_inches.padded(pad_inches)
   2352     # call adjust_bbox to save only the given area
-> 2353     restore_bbox = _tight_bbox.adjust_bbox(
   2354         self.figure, bbox_inches, self.figure.canvas.fixed_dpi)
   2356     _bbox_inches_restore = (bbox_inches, restore_bbox)
   2357 else:

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/matplotlib/_tight_bbox.py:28, in adjust_bbox(fig, bbox_inches, fixed_dpi)
     26 locator = ax.get_axes_locator()
     27 if locator is not None:
---> 28     ax.apply_aspect(locator(ax, None))
     29 locator_list.append(locator)
     30 current_pos = ax.get_position(original=False).frozen()

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/mpl_toolkits/axes_grid1/inset_locator.py:73, in AnchoredLocatorBase.__call__(self, ax, renderer)
     71 def __call__(self, ax, renderer):
     72     self.axes = ax
---> 73     bbox = self.get_window_extent(renderer)
     74     px, py = self.get_offset(bbox.width, bbox.height, 0, 0, renderer)
     75     bbox_canvas = Bbox.from_bounds(px, py, bbox.width, bbox.height)

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/matplotlib/offsetbox.py:399, in OffsetBox.get_window_extent(self, renderer)
    396 def get_window_extent(self, renderer=None):
    397     # docstring inherited
    398     if renderer is None:
--> 399         renderer = self.figure._get_renderer()
    400     bbox = self.get_bbox(renderer)
    401     try:  # Some subclasses redefine get_offset to take no args.

AttributeError: 'NoneType' object has no attribute '_get_renderer'

<Figure size 800x420 with 9 Axes>

Examine contrasts between conditions#

Viewing the ERPs for each experimental condition is a good first step, but ultimately any ERP experiment should be designed around contrasts between different conditions. In the present study, we are interested in the difference between match and mismatch items. We predicted an N400 component, which would be larger (more negative) for mismatch than match items. The N400 tends to be largest over the midline centre/posterior region of the scalp (when using averaged-mastoid reference). So here we will plot the data at electrode Cz, which typically shows the N400 effect. We use the function mne.viz.plot_compare_evokeds(), which plots waveforms for each of a list of evoked items that you pass to it:

mne.viz.plot_compare_evokeds(evokeds, 
                             picks='Cz');

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/IPython/core/formatters.py:340, in BaseFormatter.__call__(self, obj)
    338     pass
    339 else:
--> 340     return printer(obj)
    341 # Finally look for special method names
    342 method = get_real_method(obj, self.print_method)

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/IPython/core/pylabtools.py:152, in print_figure(fig, fmt, bbox_inches, base64, **kwargs)
    149     from matplotlib.backend_bases import FigureCanvasBase
    150     FigureCanvasBase(fig)
--> 152 fig.canvas.print_figure(bytes_io, **kw)
    153 data = bytes_io.getvalue()
    154 if fmt == 'svg':

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/matplotlib/backend_bases.py:2353, in FigureCanvasBase.print_figure(self, filename, dpi, facecolor, edgecolor, orientation, format, bbox_inches, pad_inches, bbox_extra_artists, backend, **kwargs)
   2350         bbox_inches = bbox_inches.padded(pad_inches)
   2352     # call adjust_bbox to save only the given area
-> 2353     restore_bbox = _tight_bbox.adjust_bbox(
   2354         self.figure, bbox_inches, self.figure.canvas.fixed_dpi)
   2356     _bbox_inches_restore = (bbox_inches, restore_bbox)
   2357 else:

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/matplotlib/_tight_bbox.py:28, in adjust_bbox(fig, bbox_inches, fixed_dpi)
     26 locator = ax.get_axes_locator()
     27 if locator is not None:
---> 28     ax.apply_aspect(locator(ax, None))
     29 locator_list.append(locator)
     30 current_pos = ax.get_position(original=False).frozen()

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/mpl_toolkits/axes_grid1/inset_locator.py:73, in AnchoredLocatorBase.__call__(self, ax, renderer)
     71 def __call__(self, ax, renderer):
     72     self.axes = ax
---> 73     bbox = self.get_window_extent(renderer)
     74     px, py = self.get_offset(bbox.width, bbox.height, 0, 0, renderer)
     75     bbox_canvas = Bbox.from_bounds(px, py, bbox.width, bbox.height)

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/matplotlib/offsetbox.py:399, in OffsetBox.get_window_extent(self, renderer)
    396 def get_window_extent(self, renderer=None):
    397     # docstring inherited
    398     if renderer is None:
--> 399         renderer = self.figure._get_renderer()
    400     bbox = self.get_bbox(renderer)
    401     try:  # Some subclasses redefine get_offset to take no args.

AttributeError: 'NoneType' object has no attribute '_get_renderer'

<Figure size 800x600 with 2 Axes>

We can make this plot a bit nicer by adding a few things to it:

firstly we define a region of interest (ROI) that includes a set of electrodes where the N400 is expected, so that we can plot the ERP waveform averaged over these electrodes
we add the combine kwarg to tell MNE how to combine the data from these electrodes
we define dictionaries mapping condition names to colours and linestyles for out plot, to make the two conditions’ lines more distinctive
we tell MNE where to put the legend and plot of the sensor locations, to ensure these don’t overlap with the waveforms

roi = ['C3', 'Cz', 'C4', 
       'P3', 'Pz', 'P4']

color_dict = {'Match':'blue', 'Mismatch':'red'}
linestyle_dict = {'Match':'-', 'Mismatch':'--'}

mne.viz.plot_compare_evokeds(evokeds, combine='mean', 
                             legend='lower right',
                             picks=roi, show_sensors='upper right',
                             colors=color_dict,
                             linestyles=linestyle_dict);

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/IPython/core/formatters.py:340, in BaseFormatter.__call__(self, obj)
    338     pass
    339 else:
--> 340     return printer(obj)
    341 # Finally look for special method names
    342 method = get_real_method(obj, self.print_method)

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/IPython/core/pylabtools.py:152, in print_figure(fig, fmt, bbox_inches, base64, **kwargs)
    149     from matplotlib.backend_bases import FigureCanvasBase
    150     FigureCanvasBase(fig)
--> 152 fig.canvas.print_figure(bytes_io, **kw)
    153 data = bytes_io.getvalue()
    154 if fmt == 'svg':

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/matplotlib/backend_bases.py:2353, in FigureCanvasBase.print_figure(self, filename, dpi, facecolor, edgecolor, orientation, format, bbox_inches, pad_inches, bbox_extra_artists, backend, **kwargs)
   2350         bbox_inches = bbox_inches.padded(pad_inches)
   2352     # call adjust_bbox to save only the given area
-> 2353     restore_bbox = _tight_bbox.adjust_bbox(
   2354         self.figure, bbox_inches, self.figure.canvas.fixed_dpi)
   2356     _bbox_inches_restore = (bbox_inches, restore_bbox)
   2357 else:

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/matplotlib/_tight_bbox.py:28, in adjust_bbox(fig, bbox_inches, fixed_dpi)
     26 locator = ax.get_axes_locator()
     27 if locator is not None:
---> 28     ax.apply_aspect(locator(ax, None))
     29 locator_list.append(locator)
     30 current_pos = ax.get_position(original=False).frozen()

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/mpl_toolkits/axes_grid1/inset_locator.py:73, in AnchoredLocatorBase.__call__(self, ax, renderer)
     71 def __call__(self, ax, renderer):
     72     self.axes = ax
---> 73     bbox = self.get_window_extent(renderer)
     74     px, py = self.get_offset(bbox.width, bbox.height, 0, 0, renderer)
     75     bbox_canvas = Bbox.from_bounds(px, py, bbox.width, bbox.height)

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/matplotlib/offsetbox.py:399, in OffsetBox.get_window_extent(self, renderer)
    396 def get_window_extent(self, renderer=None):
    397     # docstring inherited
    398     if renderer is None:
--> 399         renderer = self.figure._get_renderer()
    400     bbox = self.get_bbox(renderer)
    401     try:  # Some subclasses redefine get_offset to take no args.

AttributeError: 'NoneType' object has no attribute '_get_renderer'

<Figure size 800x600 with 2 Axes>

The N400 can be observed as the difference between the two conditions’ waveforms starting just after 400 ms. As predicted, the Mismatch waveform is more negative than the Match waveform. In most cases, the N400 lasts for about 200 ms. However, in this particular individual’s data, under these task conditions, the larger negativity for Mismatch lasts until the end of the epoch.

Saved Evoked to file#

We can save Evoked objects to files using mne.write_evokeds(). Note that this function does not accept a dictionary as input, only a singe Evoked object, or a list of Evoked objects, so below we convert the values of evokeds to a list using the list() function.

One potential source of error here is that, unlike dictionaries, lists don’t have labels for their entries. So in a list of Evoked objects, we risk not knowing later which list item corresponds to which condition. Fortunately, MNE provides a way to label Evoked objects: the .comment property:

evokeds['Match'].comment

'0.25 × Match/A + 0.25 × Match/B + 0.25 × Match/C + 0.25 × Match/D'

At the moment, this does indicate what condition the data are from, but in a somewhat messy way. Indeed, the comment there reflects the result of averaging the 4 different Match conditions into one Evoked object; the 02.5 values reflect the weighting of each condition in the average, which was based on the number of trials in each. In this case, we had equal numbers of trials of each of the four Match conditions, so all the weights are 25%. Since this information isn’t terribly relevant now that we are treating all four conditions as the same (simply Match), so we can change the contents of the .comment property through assignment:

for condition in evokeds.keys():
    evokeds[condition].comment = condition

evokeds

{'Match': <Evoked | 'Match' (average, N=72), -0.2 – 1 s, baseline -0.2 – 0 s, 16 ch, ~105 kB>,
 'Mismatch': <Evoked | 'Mismatch' (average, N=72), -0.2 – 1 s, baseline -0.2 – 0 s, 16 ch, ~105 kB>}

Now that we have comments that will help clearly identify each Evoked object later, we can write these out to a file.

MNE imposes file naming conventions (rules), which specify that Evoked file names should always end in -ave.fif. It’s good practice to use the participant ID and condition name in the file name as well. The write_evoked() function requires the output file name as its first argument, and the Evoked data as the second argument.

mne.write_evokeds('data/' + p_id + '-ave.fif', 
                  list(evokeds.values())
                 )

---------------------------------------------------------------------------
FileExistsError                           Traceback (most recent call last)
Cell In[8], line 1
----> 1 mne.write_evokeds('data/' + p_id + '-ave.fif', 
      2                   list(evokeds.values())
      3                  )

File <decorator-gen-265>:12, in write_evokeds(fname, evoked, on_mismatch, overwrite, verbose)

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/mne/evoked.py:1845, in write_evokeds(fname, evoked, on_mismatch, overwrite, verbose)
   1814 @verbose
   1815 def write_evokeds(fname, evoked, *, on_mismatch="raise", overwrite=False, verbose=None):
   1816     """Write an evoked dataset to a file.
   1817 
   1818     Parameters
   (...)
   1843         via `mne.read_evokeds`.
   1844     """
-> 1845     _write_evokeds(fname, evoked, on_mismatch=on_mismatch, overwrite=overwrite)

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/mne/evoked.py:1852, in _write_evokeds(fname, evoked, check, on_mismatch, overwrite)
   1849 """Write evoked data."""
   1850 from .dipole import DipoleFixed  # avoid circular import
-> 1852 fname = _check_fname(fname=fname, overwrite=overwrite)
   1853 if check:
   1854     check_fname(
   1855         fname, "evoked", ("-ave.fif", "-ave.fif.gz", "_ave.fif", "_ave.fif.gz")
   1856     )

File <decorator-gen-0>:12, in _check_fname(fname, overwrite, must_exist, name, need_dir, verbose)

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/mne/utils/check.py:230, in _check_fname(fname, overwrite, must_exist, name, need_dir, verbose)
    228 if fname.exists():
    229     if not overwrite:
--> 230         raise FileExistsError(
    231             "Destination file exists. Please use option "
    232             '"overwrite=True" to force overwriting.'
    233         )
    234     elif overwrite != "read":
    235         logger.info("Overwriting existing file.")

FileExistsError: Destination file exists. Please use option "overwrite=True" to force overwriting.

Create difference waves#

Finally, we can create difference waves by subtracting the waveforms of our two experimental conditions. Since the N400 manifests as a larger negativity for mismatch, it is conventional to compute this difference as Mismatch - Match (otherwise the difference between the two would appear as positive, rather than negative, voltage).

Computing difference waves allows us to view the experimental effect in a quite intuitive way: zero voltage means no difference between conditions, and and deviation from the zero line in the waveform plot (or from white in the topopmaps) indicates an experimental effect.

MNE’s combine_evoked() function allows us to combine Evoked data sets in different ways. To compute a difference, we pass the weights of [1, -1] to indicate we want to subtract the second Evoked data set listed (Match) from the first (Mismatch):

evokeds_diff = mne.combine_evoked([evokeds['Mismatch'], evokeds['Match']], 
                                             weights=[1, -1])
    
evokeds_diff

Condition	Mismatch - Match
Data kind	average
Timepoints	601 samples
Channels	16 channels
Number of averaged epochs	36.0
Time range (secs)	-0.2 – 1.0
Baseline (secs)	-0.200 – 0.000 s

Having saved this as a new Evoked object, we can plot it as a waveform.

We do a few extra things to make the plot more readable:

Whereas before we simply passed the Epochs or Evoked data to the plot_compare_evokeds() function, here we pass a dictionary, where the key is the label we want to appear in the legend and the value is the Evoked data
we add a title kwarg to give the plot an informative title

mne.viz.plot_compare_evokeds({'Mismatch-Match':evokeds_diff}, 
                             picks=roi, show_sensors='upper right',
                             combine='mean',
                             title='Difference Wave');

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/IPython/core/formatters.py:340, in BaseFormatter.__call__(self, obj)
    338     pass
    339 else:
--> 340     return printer(obj)
    341 # Finally look for special method names
    342 method = get_real_method(obj, self.print_method)

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/IPython/core/pylabtools.py:152, in print_figure(fig, fmt, bbox_inches, base64, **kwargs)
    149     from matplotlib.backend_bases import FigureCanvasBase
    150     FigureCanvasBase(fig)
--> 152 fig.canvas.print_figure(bytes_io, **kw)
    153 data = bytes_io.getvalue()
    154 if fmt == 'svg':

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/matplotlib/backend_bases.py:2353, in FigureCanvasBase.print_figure(self, filename, dpi, facecolor, edgecolor, orientation, format, bbox_inches, pad_inches, bbox_extra_artists, backend, **kwargs)
   2350         bbox_inches = bbox_inches.padded(pad_inches)
   2352     # call adjust_bbox to save only the given area
-> 2353     restore_bbox = _tight_bbox.adjust_bbox(
   2354         self.figure, bbox_inches, self.figure.canvas.fixed_dpi)
   2356     _bbox_inches_restore = (bbox_inches, restore_bbox)
   2357 else:

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/matplotlib/_tight_bbox.py:28, in adjust_bbox(fig, bbox_inches, fixed_dpi)
     26 locator = ax.get_axes_locator()
     27 if locator is not None:
---> 28     ax.apply_aspect(locator(ax, None))
     29 locator_list.append(locator)
     30 current_pos = ax.get_position(original=False).frozen()

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/mpl_toolkits/axes_grid1/inset_locator.py:73, in AnchoredLocatorBase.__call__(self, ax, renderer)
     71 def __call__(self, ax, renderer):
     72     self.axes = ax
---> 73     bbox = self.get_window_extent(renderer)
     74     px, py = self.get_offset(bbox.width, bbox.height, 0, 0, renderer)
     75     bbox_canvas = Bbox.from_bounds(px, py, bbox.width, bbox.height)

File ~/mambaforge/envs/ncil/lib/python3.10/site-packages/matplotlib/offsetbox.py:399, in OffsetBox.get_window_extent(self, renderer)
    396 def get_window_extent(self, renderer=None):
    397     # docstring inherited
    398     if renderer is None:
--> 399         renderer = self.figure._get_renderer()
    400     bbox = self.get_bbox(renderer)
    401     try:  # Some subclasses redefine get_offset to take no args.

AttributeError: 'NoneType' object has no attribute '_get_renderer'

<Figure size 800x600 with 2 Axes>

One advantage of difference waves is that they make it very clear when differences are occurring between conditions, and how large they are. When we compare waveforms for two conditions, the shape of the waveforms can sometimes make it hard to mentally compute the size of the difference - especially at time points when the waveforms’ amplitudes are changing. In a difference waveform plot, the zero line clearly indicates no difference between conditions, and so the magnitude of the difference — as well as its timing — is easy to see.

And we can plot the scalp distribution of the differences as well, which nicely shows the scalp distribution of the N400:

import numpy as np
# Specify times to plot at, as [min],[max],[stepsize]
times = np.arange(0, evokeds_diff.tmax, 0.1)

evokeds_diff.plot_topomap(times=times, average=0.050, title='Mismatch-Match');

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[11], line 5
      2 # Specify times to plot at, as [min],[max],[stepsize]
      3 times = np.arange(0, evokeds_diff.tmax, 0.1)
----> 5 evokeds_diff.plot_topomap(times=times, average=0.050, title='Mismatch-Match');

TypeError: Evoked.plot_topomap() got an unexpected keyword argument 'title'

Saving figures to files#

Since MNE uses Matplotlib for its plotting duties, it is easy to save any MNE figure as an image file. The trick to doing this is to run the plotting command with the show=False kwarg, and then running plt.savefig(). The figure will still show in your notebook, but it will also be saved to the file name you specify. The extension you give the filename determines the image file format (e.g., .png, .jpg, .pdf, etc.)

import matplotlib.pyplot as plt
evokeds_diff.plot_topomap(times=times, average=0.050, title='Mismatch-Match',
                          show=False)

plt.savefig('diffwave_topomaps.png')

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[12], line 2
      1 import matplotlib.pyplot as plt
----> 2 evokeds_diff.plot_topomap(times=times, average=0.050, title='Mismatch-Match',
      3                           show=False)
      5 plt.savefig('diffwave_topomaps.png')

TypeError: Evoked.plot_topomap() got an unexpected keyword argument 'title'

Summary#

Preprocessing EEG data and deriving ERPs is a multi-step process. It’s important to understand what each step does, why it’s applied, and how to choose the correct parameters.

There are a few additional steps that could be included in a preprocessing pipeline, that we have omitted here. For example, sometimes there there are bad channels — electrodes that had a poor connection throughout the recording, or were broken. There are ways to manually or automatically identify these and have them ignored throughout the preprocessing stream, and then interpolate data for these electrodes after artifact correction.

After each individual participant’s data has been preprocessed through a pipeline line this, the next step in analyzing an ERP experiment would be a group-level analysis. In a group-level analysis, the data from all participants would typically be averaged, and visualized using waveform plots and scalp topographic maps similar to the ones we’ve created here for one participant.

As well, statistics can be applied. An important question in applying statistics to EEG data is how to handle time - do we perform statistical comparisons between conditions at every time point, or average over periods of time when ERP components occur, or include time as an additional variable in an analysis?

One of the most common ways for ERP studies to run statistics is to average the data at each channel over a time period appropriate for the component of interest. For example, in this study a mean amplitude could be computed over the 600–800 ms time window, and then a t-test could be performed to compare the mean amplitude values between conditions.