Spatiotemporal cluster permutation on evoked data in a within subjects design

✉️ Mailing List Archive (read-only)
1

Hi all,

I'm trying to run spatiotemporal cluster permutation on evoked responses in
a within subjects design where each subject is observed in two conditions:
cond1 and cond2:

to do so I'm using the following:

# for 17 subjects and 100 ms time window

conditions= [cond1, cond2]
for j, cond in enumerate(conditions):
    X_tmp = np.zeros((17, 101, 102))

    for i, subject in enumerate(subjects):
        evoked = mne.read_evokeds(fname)
        grads= evoked.pick_types(meg="grad")
        rms = rms_grad(grads.data)
        X_tmp[i, :] = np.transpose(rms)

    if j == 0:
        X = X_tmp
    else:
        X = np.vstack((X, X_tmp))

connectivity, ch_names = find_ch_connectivity(evoked.info, ch_type='mag')

T_obs, clusters, p_values, _ = spatio_temporal_cluster_test([X[:17],
X[17:]],
                                                    n_permutations=5000,
                                                    tail=0,
                                                    n_jobs=2,

connectivity=connectivity)

After around 20 mints of running the script, the returned p_values look
like this:

p_values
Out[68]:
array([ 0.9476, 1. , 1. , 0.8854, 1. , 1. , 0.9966,
        1. , 1. , 1. , 1. , 1. , 0.9974, 1. ,
        1. , 1. , 1. , 1. ])

My question is, given the input data: [X[:17], X[17:]] where:
X[17:] contains the data observed in cond1 from all 17 subjects and
similarly X[17:] is the data observed in cond2.

are the permutations in this case done between cond1 and cond2 within the
same subject (as should be) or randomly between subjects?

Many thanks in advance

Rasha

<https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
Virus-free.
www.avast.com
<https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
<#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20180117/bde1bbe7/attachment.html

2

Hi Rasha,

1. I'm not sure I follow your script. We don't have an rms_grad function,
but I guess you're taking the root mean square across gradiometers. If so,
the shape of X_tmp should be 2D (17 x n_times) not 3D (17 * 101 * 102 in
your script), shouldn't it?

2. I think the correct way to do the within subject analysis you mention is
to use mne.stats.spatio_temporal_cluster_1samp_test: i.e. first you
subtract the two condition within subject, second you run the permutations
spatio_temporal_cluster_1samp_test(X[:17]-X[17:])

Hope that helps,

Jean-R?mi

3

Dear Jean,
many thanks for your reply.

In fact *rms_gard* is the same as *_merge_grad_data* function imported from
*mne.channels.layout* (*from mne.channels.layout import _merge_grad_data as
rms_grad*). It merges the data of each pair of grads using the RMS and
returns an array of shape (n_grads/2, n_times) so for the 204 grads I have
102 arrays and the X_temp has this shape (n_subjects, n_times, n_grads/2).

Regarding your suggesting of subtracting the conditions first and then
using mne.stats.spatio_temporal_cluster_1samp_test instead, could you
please try to explain a bit more the logic behind using this method instead
of the way I'm currently using now?

with best regards
Rasha

<https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
Virus-free.
www.avast.com
<https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
<#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>

4

In fact *rms_gard* is the same as *_merge_grad_data* function imported
from *mne.channels.layout* (*from mne.channels.layout import
_merge_grad_data as rms_grad*). It merges the data of each pair of grads
using the RMS and returns an array of shape (n_grads/2, n_times) so for the
204 grads I have 102 arrays and the X_temp has this shape (n_subjects,
n_times, n_grads/2).

Ok, that makes sense.

Regarding your suggesting of subtracting the conditions first and then
using mne.stats.spatio_temporal_cluster_1samp_test instead, could you
please try to explain a bit more the logic behind using this method instead
of the way I'm currently using now?

IIRC the 1 sample cluster test permutes the sign (of the subtraction
between the two conditions) across subjects. The test thus assesses the
probability that your distribution is distinct from the distribution based
on permuted signs.

Hope that helps,

Jean-R?mi

with best regards
Rasha

<Safe emailing Mac | Avast; Virus-free.
www.avast.com
<Safe emailing Mac | Avast;
<#m_8627169073257330069_DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>

Hi Rasha,

1. I'm not sure I follow your script. We don't have an rms_grad function,
but I guess you're taking the root mean square across gradiometers. If so,
the shape of X_tmp should be 2D (17 x n_times) not 3D (17 * 101 * 102 in
your script), shouldn't it?

2. I think the correct way to do the within subject analysis you mention
is to use mne.stats.spatio_temporal_cluster_1samp_test: i.e. first you
subtract the two condition within subject, second you run the permutations
spatio_temporal_cluster_1samp_test(X[:17]-X[17:])

Hope that helps,

Jean-R?mi

Hi all,

I'm trying to run spatiotemporal cluster permutation on evoked responses
in a within subjects design where each subject is observed in two
conditions: cond1 and cond2:

to do so I'm using the following:

# for 17 subjects and 100 ms time window

conditions= [cond1, cond2]
for j, cond in enumerate(conditions):
    X_tmp = np.zeros((17, 101, 102))

    for i, subject in enumerate(subjects):
        evoked = mne.read_evokeds(fname)
        grads= evoked.pick_types(meg="grad")
        rms = rms_grad(grads.data)
        X_tmp[i, :] = np.transpose(rms)

    if j == 0:
        X = X_tmp
    else:
        X = np.vstack((X, X_tmp))

connectivity, ch_names = find_ch_connectivity(evoked.info,
ch_type='mag')

T_obs, clusters, p_values, _ = spatio_temporal_cluster_test([X[:17],
X[17:]],
                                                    n_permutations=5000,
                                                    tail=0,
                                                    n_jobs=2,

connectivity=connectivity)

After around 20 mints of running the script, the returned p_values look
like this:

p_values
Out[68]:
array([ 0.9476, 1. , 1. , 0.8854, 1. , 1. , 0.9966,
        1. , 1. , 1. , 1. , 1. , 0.9974, 1. ,
        1. , 1. , 1. , 1. ])

My question is, given the input data: [X[:17], X[17:]] where:
X[17:] contains the data observed in cond1 from all 17 subjects and
similarly X[17:] is the data observed in cond2.

are the permutations in this case done between cond1 and cond2 within
the same subject (as should be) or randomly between subjects?

Many thanks in advance

Rasha

<Safe emailing Mac | Avast; Virus-free.
www.avast.com
<Safe emailing Mac | Avast;
<#m_8627169073257330069_m_-3461909942541472768_m_-8311896079980075393_DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>

_______________________________________________
Mne_analysis mailing list
Mne_analysis at nmr.mgh.harvard.edu
Mne_analysis Info Page

The information in this e-mail is intended only for the person to whom
it is
addressed. If you believe this e-mail was sent to you in error and the
e-mail
contains patient information, please contact the Partners Compliance
HelpLine at
MyComplianceReport.com: Compliance and Ethics Reporting . If the e-mail was sent to you
in error
but does not contain patient information, please contact the sender and
properly
dispose of the e-mail.

_______________________________________________
Mne_analysis mailing list
Mne_analysis at nmr.mgh.harvard.edu
Mne_analysis Info Page

The information in this e-mail is intended only for the person to whom it
is
addressed. If you believe this e-mail was sent to you in error and the
e-mail
contains patient information, please contact the Partners Compliance
HelpLine at
MyComplianceReport.com: Compliance and Ethics Reporting . If the e-mail was sent to you
in error
but does not contain patient information, please contact the sender and
properly
dispose of the e-mail.

_______________________________________________
Mne_analysis mailing list
Mne_analysis at nmr.mgh.harvard.edu
Mne_analysis Info Page

The information in this e-mail is intended only for the person to whom it
is
addressed. If you believe this e-mail was sent to you in error and the
e-mail
contains patient information, please contact the Partners Compliance
HelpLine at
MyComplianceReport.com: Compliance and Ethics Reporting . If the e-mail was sent to you in
error
but does not contain patient information, please contact the sender and
properly
dispose of the e-mail.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20180117/6de6fd2b/attachment-0001.html

5

Thanks a lot Jean for your clarification. I will try doing it this way.

All the best
Rasha