donders-institute / infant-cluster-effectsize Goto Github PK

although you don't have any subject details yet, I suggest you start adding subect details files anyway. that would be around here

https://github.com/DidiLamers/BeeG-analysis/blob/a4e2971ae90f264fab94ab3f38e302e0ef6410df/do_singlesubject_analysis.m#L33

The files would be called K0109.m or sub-K0109.m, would be included as code in this repo, and in the singlesubject analysis script you would simply evaluate them similar to https://github.com/robertoostenveld/Wakeman-and-Henson-2015

it now contains

{
	"Authors": [
		"Ezgi Kayhan",
		"Marlene Meyer",
		"Jill X O'Reilly",
		"Sabine Hunnius",
		"Harold Bekkering"
	],
	"BIDSVersion": "1.2.0",
	"DatasetType": "raw",
	"License": "ODC-ODbL-1.0",
	"Name": "Nine-month-old infants update their predictive models of a changing environment"
}

I think it is fine that it refers to the original acquisition and paper, but should also contain a DOI to the original paper, to the place where the data is stored (on the Donders Repository), and possibly a reference to the manuscript that we are now making.

I'd like to re-consider the way we count the repetitions of the bees. at the moment, we re-start the count after an oddball stimulus (ignoring in the count any bees that occur after the cue, like the update/no-update bees). However, since the same stimulus continues after no-update cues, we might need to take that into account.

Either way, we need to change in the script where we determine the repetitions --> we need to move it from after artefact rejection to before, otherwise we do not take into account artifacted trials although infants did see/hear those trials. I can do that once we decided which counting system we prefer

https://github.com/DidiLamers/BeeG-analysis/blob/896585028a68efd285aa596cdc305356583ed996/details_subK0109.m#L27

deze sub_xxx.m code zou eigenlijk niets moeten doen, alleen maar een aantal variabelen definiëren.

We now interpolate missing/bad channels after ICA but after that is another step during which we may exclude channels. I think we would interpolate any additional channels, rejected after that again?! Otherwise, now we end up with some participants having all NANs in the timelock data.

it is now part of the single subject script

in do_group_analysis:
I wonder whether participants should be excluded based on the percentage of trials we reject or rather the number of trials we keep. Since unlike adult experiments, not all infants will have watched all trials, the percentage value might be deceiving. Or should we at least build in a minimum number of trials per condition?

and add the results, bidsresults, and scripts directories

also here I recommend using readtable. Subsequently you can use the table syntax, i.e.participants.participant_id and you don't have to think about which column is which. This stays closer to the idea of tables (rather than matrices) and "data dictionary" json sidecar files.

https://github.com/DidiLamers/BeeG-analysis/blob/a4e2971ae90f264fab94ab3f38e302e0ef6410df/analyze.m#L27

We could even consider implementing a function (in fieldtrip) to read a tsv together with its json sidecar (if available) and then use the json to fill the Variable and Custom Metadata according to https://nl.mathworks.com/help/matlab/ref/table.html. That would be useful for the participants, but also for electrodes, channels, and other tables. I was actually just working on fieldtrip/fieldtrip#1616 and fieldtrip/fieldtrip#1601 where that would also be handy.

I propose that this block of code goes to a more general place.

https://github.com/DidiLamers/BeeG-analysis/blob/896585028a68efd285aa596cdc305356583ed996/details_subK0109.m#L14

it would be fine with me to have a general_settings.m script

that would put it (also alphabetically) in line with the do_singlesubject_analysis and do_group_analysis.

I propose that you use MATLAB readtable instead of this. Just to further standardize the code and reduce depedencies.

https://github.com/DidiLamers/BeeG-analysis/blob/a4e2971ae90f264fab94ab3f38e302e0ef6410df/do_singlesubject_analysis.m#L41

I don't think this applies

https://github.com/DidiLamers/BeeG-analysis/blob/a4e2971ae90f264fab94ab3f38e302e0ef6410df/license.txt#L1

We hadden het over BIDS in de MEG meeting. @BobBramson is geïnteresseerd te zien hoe jij het hebt gedaan met de trialfun en data2bids. Wil jij hem toegang geven tot deze repo?