[Mne_analysis] Resampling epochs

[Alex Murphy]

        External Email - Use Caution        

Hi there,

I just had a doubt I wanted to resolve before I felt fully comfortable with
an analysis I am doing. I record data at 1,000 Hz and then perform some
preprocessing and then epoch the data. I can't downsample to 100 Hz
straight away because a lot of my triggers are close together and MNE
doesn't like that (throwing errors such as multiple time instances per
sample and time indices not unique errors, even with shortest_event=0 and
the like).

Anyway, after epoching I do some manipulations of the trigger values (just
converting the incremental integer target values to class labels according
to my experimental design) and then I want to downsample the epochs.

What I notice is that the events attribute of the epochs class doesn't
change the time samples and I guess I would have expected that, perhaps.
After copying the epochs and resampling it, I have the original and my
resampled epoch data structure. If I run the following code, the following
results are shown:

print(epochs_resampled.events[-1])
print(epochs.events[-1])

[1901947       0     504]
[1901947       0     504]

That is what confused me. Basically, if MNE just looks at the index of the
third column when dealing with the epoch events from now on, I guess that's
all fine. I basically lose access to the original data so I am hoping it's
just the case that the time stamp in the first column doesn't matter
anymore.

If I save the epochs data to a file and reload it, I just want the
confirmation that absolutely no manipulation I would typically perform on
epoched data will be using the first column. It is effectively meaningless
after epoching, right?

Kind regards
Alex
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20181203/afd6ba0d/attachment.html