Using mne.event.merge_events works fine when I was just creating one new event (as does the alternative method of combining events at the evoked stage) but I'm unsure of the most sensible way to proceed given I want to create multiple new events. I originally did this:
but then, because I'm not replacing events, I end up with multiple copies of all events (even those not included in the merge_events function) at a given sample point, which isn't ideal.
As a new MNE Python user (I previously used MNE) I would be grateful for any advice as the best way to do this.
It seems what you want to do is to use the hierarchical epochs selection.
If you have an epochs object which contains condA, condB, condC and condD,
you can get epochs of condE by simply using epochs['condA', 'condB'].
Please take a look at the MNE reference API for the epochs object: http://martinos.org/mne/stable/generated/mne.Epochs.html#mne.Epochs (look
at the Notes section). Does this solve your problem?
Many thanks for the reply and the pointer. This seems helpful to access different combinations of the events, but is it then possible to create multiple new epoch objects (condE and condF) which can be then saved together in a single evoked file along with the originals? Ideally I would have the event/condition averages for all conditions (new and old) in one file.
Maybe I misunderstand what you are trying to do, but it seems
like mne.epochs.combine_event_ids might be helpful. I'm not sure what you
mean about saving them all into one file but you could define new event ids
to combine epochs across conditions then create evoked files from those.
Thank you for the suggestion. My understanding of mne.epochs.combine_event_ids is that it replaces the original events with new ones whereas I would like to retain the original events which is why I initially tried mne.event.merge_events.
I can now do what I want to do using Mainak?s suggestion to use the hierarchical feature of epoch selection. I created multiple evoked files corresponding to my various event combinations then saved the combinations into one file using mne.write_evokeds.
This might help, there's a argument with merge_events named
'replace_events' that allows you to retain the original values when set to
False. In your example, you could combine 11 and 12 and assign to a value
like 99. In your event_id dict, you could then assign CondE to 99.
Thank you. Your suggestion is what I?d originally hoped to do but I would like *multiple* new combinations (e.g., condE(11,12) *and* condF(13,14)) and I couldn?t figure out how I could use the function to do this to create a list of the events.
Initially I did:
events=mne.event.merge_events(events, [11,12], 1112, replace_events=False)
events=mne.event.merge_events(events, [13,14], 1314 replace_events=False)
In doing mne.event.merge_events the second time, at a given sample point the original and new event codes are repeated twice (event codes not included in recombinations are repeated 4 times).
Was any resolution ever found on this? I'm now experiencing the same thing where trying to multiple merge_events before epoching results in an obscene amount of trial counts per condition (e.g. my condition 1 nave goes from 69 > 228), so I believe my problem to identical to Lucy's below. Thanks for any help or input.
Sorry I followed up directly but forgot to relay it back here. I made pull
request to fix this behavior. If you would like to follow along the
process, you can do so here: https://github.com/mne-tools/mne-python/pull/3015
Just to follow up here, the merge event bug has been resolved
<https://github.com/mne-tools/mne-python/pull/3076> and it is now available
on the dev version. This fix will be included in the v0.12 release in the
nearby future.