remigau.bsky.social
I used to try to understand how the brain works.
Now I tell people how to name their files and variables.
94 posts
723 followers
742 following
Regular Contributor
Active Commenter
comment in response to
post
Will check it out as I keep asking people: show me it can be done!
Though I have also heard people being stopped from sharing group level stat maps because they were told it could lead to individuals being reindentified. And that sounds wild to me.
comment in response to
post
Knee jerk reaction: for the plot showing there use a proper sequential colormap and ditch the rainbow colormap (except maybe on #LGBTQA pride month) and add some isoline on the plot.
comment in response to
post
Or... We need to ask less from individual scientists, promote more team work and hire more research software engineer who would be more in charge of the coding... But things will get way worse before any of that happens... If they ever do.
comment in response to
post
Given the proportion of scientific code with regularly run test suite, I have come to pretty much distrust almost everything.
Given the AI assisted code contributions I have seen on some open source scientific software, I would say that teaching students how to code is even more important now.
comment in response to
post
My hunch is: the latter (once again just a wet finger guestimation).
The theoretical framing for a lot fmri research is so thin that I often wonder if the cost of the study is justified.
comment in response to
post
Maybe I am being super negative but I see this lack of full understanding of all the methods used any scientist as the norm rather than the exception. This is just based on my personal non-systematic experience in biology, psychology, neuroscience. Happy to see any data on this.
comment in response to
post
If it helps anyone: I have been working on a "BIDS inventory" of ALL (not just openneuro) open BIDS datalad datasets to gather info about them (like 'do they have accessible derivatives?').
Currently it lives here as a disgustingly-flat lists.
github.com/neurodatasci...
github.com/neurodatasci...
comment in response to
post
All those derivatives are generated by openneuro and also all accessible / browsable on that github organisation: github.com/OpenNeuroDer...
comment in response to
post
Agreed there is no easy way to search for datasets with derivatives on openneuro.
You will know if the dataset you are looking has some if there there is a "derivatives" tab on the landing page of a given dataset.
For example:
openneuro.org/datasets/ds0...
comment in response to
post
Also not sure how trendy projective tests are clinical psych nowadays, but at the time in France I only ever heard about those.
Still remember a TA saying that the Rorschach test was the best way to understand a patient...
comment in response to
post
Oh it is clearly interesting from that angle but in a clinical psych curriculum... Not so much...
Also the content I was exposed to was very uncritical of any of the theories...
comment in response to
post
Well I don't know about other countries but a lot of psychoanalysis could be removed from the psych curriculum of many French universities.
comment in response to
post
@kordinglab.bsky.social
Why that dataset specifically?
If it's for a demo you got oodles of preprocessed dataset on openneuro.
comment in response to
post
Ha.
Neuroscout runs study on that dataset. Neuroscout uses fitlins and the bids stats model and requires an fmriprep derivative...
So...
Voilà .
github.com/neuroscout-d...
comment in response to
post
I call that the spice of life...
Once the talk is done.
comment in response to
post
Well the dataset is there: ds000113
But there is no associated derivatives on openneuro.
comment in response to
post
Only started enjoying giving talks when I could do them in my own style.
Now, because I mostly do software dev, I just enjoy doing live demos just so the crowd gets the chance to see me fail live.
comment in response to
post
Would defo help to have a public fmriprep output of the dataset. Do you know if there is one.
comment in response to
post
That website is not mobile friendly. 😜
comment in response to
post
Actually, with nilearn, from fmriprep outputs it is really fast to try different strategies for preprocessing and picking different atlases. Check https://main-educational.github.io/intro_nilearn/functional-connectivity-with-nilearn.html
comment in response to
post
> Peer review, which has for centuries been the standard tool to determine an academic paper’s suitability for publication,
Centuries???!!! They meant decades, right?