Today's Psych-DS highlight is folder structure! On the one hand, Psych-DS is extremely permissive compared to other standards what's inside data/ - flat list of files, subfolders📂📂📂, whatever. On the other it's a shift from many people's current workflows - if it's NOT data, get it out of there!! - ThreadSky

mekline.bsky.social • 22 days ago

Today's Psych-DS highlight is folder structure!

On the one hand, Psych-DS is extremely permissive compared to other standards what's inside data/ - flat list of files, subfolders📂📂📂, whatever.

On the other it's a shift from many people's current workflows - if it's NOT data, get it out of there!!

Reposted from Melissa Kline Struhl

Psych-DS is (1) spellcheck for your datasets and (2) a pathway to standardizing data in our academic fields that *everyone* can learn.

And it's live RIGHT NOW!

psych-ds.github.io

(This is the announcement post I've been leading up to)

Comments

cghlewis.bsky.social•22 days ago

I second this, mainly because I view 2 different types of data processes, especially in large, multi-year projects. There is the process of storing raw data, along with a general clean version. And then there is the process where people take that "clean" data, and further modify it for analysis.

cghlewis.bsky.social•22 days ago

So I think structures like this make the most sense for the researchers I often work with. It doesn't have to be this exactly. Just something along these lines.

mekline.bsky.social•22 days ago

To use Psych-DS successfully, you need to adopt a common practice that's mentioned in basically all the data management resources I've been sharing - Designate a single project folder, and then inside it, group your files by type. Roughly like this:

https://bsky.app/profile/mekline.bsky.social/post/3llu6p2ysjc2a/quotes

mekline.bsky.social•22 days ago

This is a great example of how planning for open science, even if you aren't sure what you'll share, yields big improvements in your personal workflows - if you do this early, you'll have a better time doing your science, AND be ready to share (w/collaborators, or the world) without too much prep.

mekline.bsky.social•22 days ago

There are roughly three big Qs/objections I've gotten to this approach, all of which I think are handle-able - but I'm curious what people think!

1 - What about files not on my personal machine
2 - How do I get my data to talk to my analysis if they are separate?
3 - But multiple experiments

mekline.bsky.social•22 days ago

(1) - Let's say you have some raw data files stored on a special server somewhere (or in some other way you don't control your original data source), and then you create annotation data that you keep locally.

That's fine! Make the folder structure anyway, and make 'breadcrumbs' for a user to follow

mekline.bsky.social•22 days ago

Here's a talk I gave that covers this, in the context of projects w/ sensitive/PII data alongside more shareable data:
https://docs.google.com/presentation/d/1ElpmktJUlxnaCoamLBBuuxq_wndfElXlBjoMMbH8VoA/edit?usp=sharing

Roughly, you want a document (a README file! Story for another day) that explains *where* that stuff is and explains how to get it for analysis

mekline.bsky.social•22 days ago

(2) This one kept me from adopting separate data/analysis folders for a (in retrospect) surprising amount of time early in grad school, so somewhat selfishly I think this is a great example of the kinds of real but easy-to-neglect barriers that keep people from following common recommendations.

Comments

Posting Rules

Reply