Thoughts on displaying scatter plot data

Visual Design
#1

Hi all, I was wondering what your opinion was on this data visualisation issue I was having. I’ve created a scatter plot to check for edge cases in disease dispersion in the united states using data from the Tycho Project.

I’ve been looking through this reference material for ideas: https://www.oreilly.com/library/view/designing-data-visualizations/9781449314774/ch04.html

My current design has mapped ‘state’ to both ‘size’ and ‘shape’ using Tableau, as I think this makes it easier and clearer to visualise which state is where on the graph. However I’m not sure if it’s bad form to map state to both size and shape or if I should just choose one of them. Here’s my current scatter plot, any advice is greatly appreciated!

image
image760×452 51.4 KB

#2

Hi gregmt99,

I can understand your reasoning. You want to make it easy for people to find a state and I like the idea of using shape as a legend.

However, you’re clearly running into issues here.

  • First, you have my symbols and more symbols become hard to discriminate (50).
  • The second thing I am seeing from your legend is that some symbols are used more than once (the diamond is used for CO and IN and there will be more down the list. This is because Tableau run out of shapes! :slight_smile:
  • Third, searching for shapes in the scatterplot is still a tricky task, given the 50 symbols.
  • As for size, I understand your goal but I do not think it works. The differences between sizes are too small to be noticed.

I am admitting I am not a master in Tableau, but here are some solutions (and maybe other tools can do the trick)

  • I’d try to put the state abbreviations into the scatterplot. They are quite small.
  • If that doesn’t work, label some states only?
  • you could also color the states according to some grouping, such as east, north, south, west or just alphabetically. and then reuse the symbols. Using color+shape is easier than size and shape.
  • Orientation is also a nice variable that often is underused, I find. However, I doubt that humans can discriminate 50 orientations…

That would be my 5cents for now. Keep some iterations posted.
Hope that helps!

cheers,
ben