Vertical histograms FTW
Posted in Uncategorized
I might be alone in this, and I’m willing to die on this hill, but here it is:
Vertical histograms are better than all of these: violin plots, box plots, kernel density plots, beeswarm plots. (if you don’t know those terms, simply do a web search for them for examples)
Just show me the naked distribution. Don’t try to mirror it around a y-axis, or smooth it, or any other nonsense. Just show me the naked histogram.
This example is from this paper and Leah did an excellent job on the figures.
I just wanted to comment that, I think bee swarm plots are great for n ? 100. They not only allow you to see the distribution, but also to see the Ns. Obviously, with large Ns, beeswarms become less useful.