Allow automated metadata generation to be bounded by "row events" instead of explicit time windows #12118
Conversation
…tead of explicit time windows
hoechenberger
force-pushed
the
metadata
branch
from
236a756
to
6831421
Compare
hoechenberger
requested review from
larsoner,
drammock,
agramfort,
dengemann and
mscheltienne
as code owners
| """Test that make_metadata works.""" | ||
| pytest.importorskip("pandas") | ||
| raw, all_events, _ = _get_data() | ||
| tmin, tmax = -0.5, 1.5 | ||
| sfreq = raw.info["sfreq"] | ||
| kwargs = dict( | ||
| events=all_events, |
There was a problem hiding this comment.
It would be great in principle to add this to an example or tutorial, but maybe we don't have an appropriate dataset (?).
In lieu of (or in addition to) that, could you add a test that shows that this does what it's supposed to do? Like create a RawArray and events that have uneven spacing, e.g., the second twice as long as the first and third, and the fourth four times as long or whatever. Then make sure that the metadata reflects... whatever it's supposed to reflect. And that creating Epochs using those events "does the right thing" -- which isn't immediately clear to me without seeing such an example in tests or in our docs.
There was a problem hiding this comment.
I agree, both the code and the tests have gotten extremely convoluted and it's difficult for me to tell what's actually happening just by looking at the code. Should probably be refactored sometime in the not-to-distant future. I will try to add a test like you proposed.
|
@larsoner I've added a test case with a slightly verbose "setup" where I first define the events in a very user-readable way, convert them to annotations, and only then create events from them. Made it easier for me to follow what I'm actually testing, and inspect intermediate results in a a human-readable way. |
|
Failing tests seem unrelated. |
|
@larsoner This seems to be working, I've been using it with a real-world dataset successfully and so far haven't encountered any obvious problems |
|
Thanks for reviewing! |
* upstream/main: (35 commits) [DOC] Add documentation for setting montage order (mne-tools#12160) Fix inferring fiducials from EEGLAB (mne-tools#12165) Try to fix ICA Report (mne-tools#12167) BUG: Fix bug with Report.add_ica component number (mne-tools#12156) MAINT: Add rstcheck to CIs and pre-commit (mne-tools#12163) DOC: fix sphinx style typos (mne-tools#12161) MAINT: Fix linkcheck (mne-tools#12162) ENH: Add multiple label support to source_band_induced_power, source_induced_power (mne-tools#12026) Allow automated metadata generation to be bounded by "row events" instead of explicit time windows (mne-tools#12118) ENH: Collapse only in doc gen (mne-tools#12145) [pre-commit.ci] pre-commit autoupdate (mne-tools#12155) BUG: Fix bug with interior points not showing (mne-tools#12148) ENH: Warn about versions in sys_info (mne-tools#12146) Fix in conftest.py (mne-tools#12150) ENH: set color for bad channel with spatial_colors=True (mne-tools#12142) DOC: Better documentation of realign_raw (mne-tools#12135) Add mne-icalabel wildcard (mne-tools#12143) Remove LGTM.com configuration file (mne-tools#12139) DOC: Fix typo found by codespell (mne-tools#12140) DOC: Document governance updates (mne-tools#12133) ...
* upstream/main: (26 commits) FIX: Fix bug with coreg scalars (mne-tools#12164) Changed casting rule in np.clip to allow reading of raw GDF files (mne-tools#12168) [DOC] Add documentation for setting montage order (mne-tools#12160) Fix inferring fiducials from EEGLAB (mne-tools#12165) Try to fix ICA Report (mne-tools#12167) BUG: Fix bug with Report.add_ica component number (mne-tools#12156) MAINT: Add rstcheck to CIs and pre-commit (mne-tools#12163) DOC: fix sphinx style typos (mne-tools#12161) MAINT: Fix linkcheck (mne-tools#12162) ENH: Add multiple label support to source_band_induced_power, source_induced_power (mne-tools#12026) Allow automated metadata generation to be bounded by "row events" instead of explicit time windows (mne-tools#12118) ENH: Collapse only in doc gen (mne-tools#12145) [pre-commit.ci] pre-commit autoupdate (mne-tools#12155) BUG: Fix bug with interior points not showing (mne-tools#12148) ENH: Warn about versions in sys_info (mne-tools#12146) Fix in conftest.py (mne-tools#12150) ENH: set color for bad channel with spatial_colors=True (mne-tools#12142) DOC: Better documentation of realign_raw (mne-tools#12135) Add mne-icalabel wildcard (mne-tools#12143) Remove LGTM.com configuration file (mne-tools#12139) ...
…tead of explicit time windows (mne-tools#12118)
This PR makes it possible to generate metadata (via
mne.epochs.make_metadata()) based on time windows of varying duration, bounded by user-specified events. Specifically, users can now passtmin=Noneandtmax=None. Then, all events occuring between the time-locked events ("row events") will be added to one row of the metadata.Use case:
I have EEG data where each trial starts with a visual cue. The experiment is self-paced, hence trial durations vary greatly within and between participants. Specifying the same rigid time window for metadata generation for all trials by passing
tminandtmaxas a float (time relative to the time-locked event, in seconds) didn't yield satifisfactory results. A much more natural approach for me was to simply add all events between two cues to the metadata.