Comments on: My Five Rules for Data Visualization

By: Elisabeth

Elisabeth — Fri, 22 Jan 2010 18:56:58 +0000

Don’t forget about red/green colorblindness. If you must use colors, do NOT include red and green in the same plot. And be consistent on what the colors represent. I use this to control colors in R:
my.palette <- colorRampPalette(c("white", "yellow", "orange", "red"), space="rgb")

Then, within whatever plotting mechanism I am using, I use:
col.regions=my.palette(20) #The 20 just refers to how shading will occur, I think.

This color scheme also works well if you have to print it in black and white. The shading makes sense.

By: Drew Conway

Drew Conway — Sat, 12 Dec 2009 01:42:03 +0000

The best resource I have seen was suggested by @dwf and is at http://colorbrewer2.org/ He actually noted that the default color scheme for ggplot2 was very bad for the colorblind, and suggested I check out this site, which has been extremely useful.

By: Nick

Nick — Sat, 12 Dec 2009 01:06:02 +0000

I found this link a while back (Hackernews?) http://www.research.ibm.com/people/l/lloydt/color/color.HTM. It has some examples that describe how color can be used to accentuate (or muddle) your visualizations. @Dugolo, it seems like it’s not just about what colors you pick, but how want you want the colors to do. Adobe’s Kuler can be a good place for inspiration, although your taste could conflict with your purpose. http://kuler.adobe.com

By: Dugolo

Dugolo — Fri, 11 Dec 2009 21:10:03 +0000

Hi Drew, thanks for the post. You write really helpful stuff!

Here’s a question for you. I agree with point #4. Do you (or anyone else) know of a tool that helps choose distinct colors? For example, I have a stacked graph that contains 10 colored areas but as much as I try, I can’t seem to choose good colors and the default matplotlib colors repeat after about 8 or so.

Seems like there should be a pallet website that can answer the question of: What would be n colors that are the most distinctive visually? For any given n.

By: Michael

Michael — Thu, 10 Dec 2009 14:51:35 +0000

Excellent post, Drew – and thanks for pointing out Security Crank to me. A new blog to follow.

But, shame on you – Sctoch “Whiskey”? If it’s Scotch, it’s “Whisky.”

By: Data Visualization « UNL Political Science Graduate Student Blog

Data Visualization « UNL Political Science Graduate Student Blog — Fri, 04 Dec 2009 16:45:35 +0000

[...] December 4, 2009 · Leave a Comment I thought this might be helpful for any and all who may be presenting at conferences in the future: The Five Rules of Data Visualization. [...]

By: Drew Conway

Drew Conway — Fri, 04 Dec 2009 14:13:59 +0000

That is an excellent point, and thanks for raising it. I actually do not know, but my suspicion is they are in USD.

By: C. Ross Jam

C. Ross Jam — Fri, 04 Dec 2009 14:07:28 +0000

Units! Will someone think of the units! Are those people tipping in dollars, pounds sterling, or bhats? Tens, hundreds, or thousands?

Enquiring minds want to know!

By: Drew Conway

Drew Conway — Fri, 04 Dec 2009 12:36:05 +0000

Goodness gracious that is ugly!

By: Ryan

Ryan — Fri, 04 Dec 2009 08:27:40 +0000

The from the New York Times just posted an item which explains, according to them, why it took so long to figure out the Afghanistan Policy: Had you posted this about 4 months ago, we'd have been 3.6 months into the new Obama Policy!