I really enjoy how raincloud plots help me see both the forest and the trees.
Summary stats consistently obscure so much, and I'm quite skeptical of model-fitting in cognition-land, so I like to have my eyes on individual data points as much as possible, but ideally "in context"
Conversation
Replying to
It's so humbling all the flailing involved in trying to make sense of messy human data. My lab notebook basically looks like a dartboard.
2
21
Replying to
This is awesome. So often data analysis critically depends on noticing artifacts of data processing (sentinel values, harmonic numbers) and the assumptions underlying any kind of summary are a great way to never notice the garbage before you GIGO.
5
Yes, but not much help unless we use maths that take the variability into account, which is mostly unheard of. A case in point being epidemiology, where all modelling paradigms, including mean-field theory, are garbage, despite empiricists producing lots of plots like these.
I use box and whiskers all the time. I really like the density plotting on that vis. What library do you use for viz?
1
1
ggplot.
3
Couldn't agree more. Devil is in the details. Wonderful plot. 😀
I play with this very often when someone gives me stddev without a histogram and/or pointcloud. autodesk.com/research/publi and in pdf form autodesk.com/content/dam/au
2
1
4
'cause everything is gaussian, right? Binomial what? Outlier whom? I like the histogram integrated in the chart very much 🙇♂️. I usually put it on the side of the point cloud, when I only display one series.
