ERP visualization -- how to plot confidence intervals/standard error of grand averaged ERP

šŸ’¬ Support & Discussions
,
1

I canā€™t seem to find a way to plot grand-averaged ERPs together with shaded confidence intervals. The closest thing I could find is here, though I donā€™t want the image map.

Can anyone give me a hint? Thanks!

1 Like
2

You can pass a dict of lists or a list of lists to plot_compare_evokeds to produce (parametric) confidence bands.

1 Like
3

Could you give an example?

4 
# Load example data

import os
import mne


sample_data_folder = mne.datasets.sample.data_path()
sample_data_raw_file = os.path.join(sample_data_folder, 'MEG', 'sample',
                                    'sample_audvis_raw.fif')
raw = (mne.io.read_raw_fif(sample_data_raw_file, verbose=False)
       .pick_types(meg=False, eeg=True, stim=True)
       .load_data()
       .filter(l_freq=None, h_freq=40))
raw.set_eeg_reference('average')

events = mne.find_events(raw)
event_id = {'auditory/left': 1,
            'auditory/right': 2,
            'visual/left': 3,
            'visual/right': 4}
epochs = mne.Epochs(raw, events, event_id=event_id, preload=True)

orig_evoked_audio = epochs['auditory'].average()
orig_evoked_visual = epochs['visual'].average()

# Simulate 5 participants

all_evoked_audio = []
all_evoked_visual = []

cov = mne.make_ad_hoc_cov(raw.info)
iir_filter = [0.2, -0.2, 0.04]

for participant in range(5):
    evoked_audio = orig_evoked_audio.copy()
    evoked_visual = orig_evoked_visual.copy()

    mne.simulation.add_noise(evoked_audio, cov=cov,
                             iir_filter=iir_filter)
    mne.simulation.add_noise(evoked_visual, cov=cov,
                             iir_filter=iir_filter)
    
    all_evoked_audio.append(evoked_audio)
    all_evoked_visual.append(evoked_visual)

# Plot evokeds with 95% CIs

# list of lists
mne.viz.plot_compare_evokeds([all_evoked_audio,
                              all_evoked_visual],
                             ci=0.95)

# dict of lists
mne.viz.plot_compare_evokeds(dict(audio=all_evoked_audio,
                                  visual=all_evoked_visual),
                             ci=0.95)
1 Like
5

Thanks, @richard!

I did a similar thing with my data, but I still donā€™t get shaded plotting of the CIsā€¦ do I also need to tweak the color and/or cmap arguments of the function?

6

No, it should work out of the box. You can try to request a different CI, e.g. 99%, by passing ci=0.99. This should make the confidence bands wider. If you still donā€™t see anything, then something is wrong with your data ā€“ maybe all evokeds are the same?

7

Oops ā€“ I mistakenly plotted the grand-averages and not the evokeds. Thank you!!

8

Quick question: how can I plot the grand average and its CIs of a single condition?

Right now, I have all by-subjects evokeds of one condition listed in a single variable. I would like to plot the grand average of all evokeds for that condition? The module mne.viz.plot_compare_evokeds() seems to parse each subject as a different condition, but what I want is that the evokeds of all subjects were averaged together.

Thanks!

9

Pass a list of your list of evokeds, or a dict of you lists of evokeds, as the first parameter of plot_compare_evokeds().
Quoting the documentation:

ā€¦ If a [dict/list] of lists, the unweighted mean is plotted as a time series and the parametric confidence interval is plotted as a shaded area. All instances must have the same shape - channel numbers, time points etc. If dict, keys must be of type str.

So you could do something like:

mne.viz.plot_compare_evokeds({ā€˜Meanā€™: list_of_evokeds}, ...)
1 Like
10

Thanks!

Two follow-up questions, if I may:

  1. Is there a way (within or outside mne.viz.plot_compare_evokeds()) to force tick labels of either axis. E.g., in my case the tick labels of the y-axis are printed by multiples of 2 (0, 2, 4, 6, etc.), but I would like to have all integers (0, 1, 2, 3, 4, etc.)

  2. I am having trouble understanding how to plot multiple channels as subplots. The documentation says:

Axes object to plot into. If plotting multiple channel types (or multiple channels when combine=None ), axes should be a list of appropriate length containing Axes objects. If 'topo' , a new Figure is created with one axis for each channel, in a topographical layout. If None , a new Figure is created for each channel type. Defaults to None .

So, I tried to do the following:

picks = ['E24', 'E36', 'E44', 'E69', 'E87']
    
axes = [plt.axes() for pick in picks]
    
mne.viz.plot_compare_evokeds({'mean': allEvokeds}, ylim=dict(eeg=[-5,5]), picks=picks,
                             title=None, axes=axes, ci=0.95, show_sensors=False, 
                             truncate_yaxis = False, truncate_xaxis=False, legend=False,
                             colors = {'mean': '#808080'}, combine=None)

But I must be doing something wrong, as no plot appears.

Thank you in advance!

11

For your second question, this part of the docstring is relevant:

If combine is None, channels are combined by computing GFP,
unless picks is a single channel (not channel type) or
axes='topo', in which cases no combining is performed.

In other words, you canā€™t pass in 5 picks and a list of 5 axes, and expect to get each channel on a separate subplot (itā€™s a reasonable thing to want to do, but itā€™s just not how the function is written).

Something like this ought to work:

picks = ['E24', 'E36', 'E44', 'E69', 'E87']
fig, axes = plt.subplots(5, 1)
for pick, ax in zip(picks, axes):
    mne.viz.plot_compare_evokeds(evoked, combine=None, picks=pick, axes=ax)
1 Like