Join 3,433 readers in helping fund MetaFilter (Hide)


Presenting Data
May 16, 2014 4:48 PM   Subscribe

A checklist for those making graphs from Stephanie Evergreen and Ann Emery. This is a useful tool for teaching scientists and others some of the rules of data presentation in graph form.
posted by sciencegeek (23 comments total) 51 users marked this as a favorite

 
Not an original thought, but we need to stop focusing on teaching geometry, algebra, and calculus to High School students. Statistics is almost unquestionably the more important survival skill in modern society, and a staggering percentage of adults are completely illiterate when it comes to stats.

Including scientists and journalists.
posted by schmod at 5:17 PM on May 16 [15 favorites]


This is really really a needed thing, but I'm disappointed to see a lot of things that seem more like matters of taste being elevated to full on rules.
posted by Kid Charlemagne at 5:25 PM on May 16 [8 favorites]


Nice find. Recently worked on a project with three different designers on the client side, and ended up with three different styles and labels in every direction. This would help.
posted by mochapickle at 5:34 PM on May 16


This is really really a needed thing, but I'm disappointed to see a lot of things that seem more like matters of taste being elevated to full on rules.

Yeah, it's a bit of a hopeful wishlist, really. Insisting on horizontal text everywhere? That's great until you need full species names on the x-axis of a community chart.
posted by Jimbob at 5:45 PM on May 16 [6 favorites]


Titles are the hardest. I'm tired of seeing graphs from students titled "Position vs. time." I can read the axes, people. Tell me something useful in the title. The prompt that I am trying get them to follow is to describe what they were doing when gathering the data on the graph. Sadly, sometimes I get "Measuring position and time."
posted by BrashTech at 6:15 PM on May 16 [1 favorite]


And
putting
your
text
like
this
for
a
complex
variable
name
will
make
me
despise
you.

I'm truly sory but it had to be done as a warning to those who come after.
posted by Kid Charlemagne at 6:22 PM on May 16 [1 favorite]


text
like
that
makes
me
wince.

I think that these are guidelines rather than rules.
posted by sciencegeek at 6:29 PM on May 16


Insisting on horizontal text everywhere? That's great until you need full species names on the x-axis of a community chart.

"Line labels and axis labels can deviate from this rule and still receive full points."

It is essentially a style guide for graphics, but assuming the data analysis is good, style's what'll kill your graphic. And I say that as someone who's made plenty of terrible graphics.
posted by brentajones at 6:38 PM on May 16


I mean if you label your axes at all you're already doing better than a lot of people. Also starting the axes at zero (i.e., avoiding truncated graphs) is nice to do.
posted by quiet coyote at 6:53 PM on May 16


Yeah, this looks like a style guide that doesn't capture the real graphing failures I've seen.

This is my favorite bad graph. It's not real data, but a cartoon about EVE Online's notoriously difficult learning curve. But ignore the dead bodies, read the axes, look at the lines, and see what the graph is really saying.
posted by ryanrs at 7:05 PM on May 16 [1 favorite]


I would have received negative comments from professors on most of these items. Just off the top of my head, some conflicting advice I've received from professors, either as grading rubric guidelines or direct feedback: "axis title should be vertical," "title of a figure goes at the bottom, title of a table goes at the top," "legend goes on the side, these labels are noisy." The people who grade my work generally don't observe these guidelines and they grade accordingly. I'm not far enough along in my career to say about journals yet, though. I'm in ecology & environmental science though, and I notice this is specifically targeted to social scientists, so maybe we just haven't caught on yet.
posted by dialetheia at 7:35 PM on May 16


__ Make color scheme legible/discernible on projectors (orthogonal: order the color key to match presentation of data categories)
posted by Blazecock Pileon at 8:53 PM on May 16


Reminds of this gif on simplifying your chart.

We've all been subjected to presentations where the message was obscured by bad design, but this check-list seems to mix "design preference" with "design effectiveness".

On the other hand, anything that decreases the amount of bad clip-art or dark grey on dark blue text is OK with me.

(Also, there really is trade group conference for everything, isn't there?)
posted by madajb at 11:42 PM on May 16 [1 favorite]


Statements like 'avoid 3D graphs' seem prohibitively narrow minded and a retardant to our aggregate societal design creativity. Some of the most interesting and information rich graphs and graphics go way beyond the paper media conventions we inherited from a guy named William Playfair. I recall he was credited with making the first line charts and had to explain how -if you took a moment to look, the picture tells a story. I remember when avatars were cool and virtual worlds held the promise of new kinds of knowledge gardens. But YouTube and cat videos won.

Most of these principles would apply to the eventual mash-ups of GPS with multivariate combinations that hold the reserve and hope of some inflectionary epiphany of collective self awareness that could be the key to saving a planet someday. After that day, saying no to 3D will be a quaint anachronism like people sitting around a radio for their access to the world. A lot could stand to change if or when big knowledge tools break out of the web of pages that ties us to text and pie charts. Big companies have been leading the way with internal tools that are kept close to their chests due to the resultant insight's pivotal role in strategic executive decisions that will make or break the twittergooglefacebooksPGExonEtcras. They use them to see us all better than we know ourselves. Dimension is just another word for a degree of freedom and once you've seen beyond the cave it's hard to go back to the still sadly hegemonic flatlands of data representation most people simply accept as the only way.

It's almost like Tufte and MSFT are leaders of a massive conspiracy to keep the rare and beautiful creative data analyst chained to conventions that keep the worlds they serve in the dark by making the very idea of innovation in n-dimensional spaces (the human imagination's inalienable and native home) some 'far out' thing. Like a visualization of how eating that hamburger while feeling self righteous about climate change deniers might mean we'd all really need to change some fundamental behaviors... forget about it. Keep information in a format where complex ideas can't be understood without reading thousands of pages and becoming an expert instead of poetically exploring how thousands of variables could be crafted into interactive narrative spaces your grandmother might understand before heading to the grocery store. I'm just saying maybe a virtual knowledge commons of some kind might be worth exploring. All that said, I love this list and will share it with a few folks at work. Thank you for posting it.
posted by astrobiophysican at 1:08 AM on May 17


Graph is two dimensional: Avoid three dimensional displays, bevels, and other distortions

I'm like the one person at my work who thinks of this as an ironclad rule, so I feel their pain, but 3D foofarah isn't going away. Hell, there's still a huge subset of the pop that thinks fucking drop shadows are the absolute beans, to the point where they're part of the default styling in multiple programs I've worked with. The picture plane is dead for now kids.
posted by Lentrohamsanin at 5:13 AM on May 17


Jean-Luc Doumont has a book on communication for scientists and engineers and produced a good handout on making effective charts.
posted by euphorb at 5:35 AM on May 17 [4 favorites]


astrobiophysican, I interpreted the remarks about no 3D as a plea to avoid useless things like 3D pie charts. There are obviously cases where response surfaces and things make a lot more sense, and convey more information, in an isometric projection. I have never encountered a situation in my research where a 3D bar chart would have been appropriate. I agree with you that it's not helpful to "ban" useful tools, but I don't think that's really what's going on here. I saw more pointless 3D graphs at the Plant & Animal Genome meetings this year than you could shake a copy of Cleveland at. Ugh.
posted by wintermind at 7:13 AM on May 17 [3 favorites]


The checklist is well meaning but something about the way it's put together lacks soul. Information display is design, is art. It's not some high complex art, it's relatively simple to do OK, but maybe it's not so simple as to boil down to a checklist in a PDF. On the other hand it's handy to have a list that says "3d pie charts in Excel are not OK".

One of my favorite things right now is the subreddit /r/dataisugly. Excellent collection of bad information visualizations. Like this fancy Eurovision voting chord diagram, the current top post. I'm not even sure where to begin with that one. Or end. It is a flat circle.
posted by Nelson at 7:20 AM on May 17


Thanks! Even if there is some subjectivity in this, a checklist that breaks down the problem and helps give some accounting rules like this is a great idea and an improvement over the present alternative. I've already added the PDF to my lecture slides for my students, and look forward to showing it to them.

The problem with data visualization is a lot of professors actually don't know this stuff either. The availability of exotic data sets happened so quickly that most of us are playing catch up. That is one way of interpreting anyway the journal of economic perspectives publishing an article devoted solely to teaching professional phd economists on better tools at displaying data.

www.aeaweb.org/articles.php?doi=10.1257/jep.28.1.209

I own tuftes books, but I feel often like they book is just making fun of me.
posted by scunning at 8:51 AM on May 17 [1 favorite]


Make color scheme legible/discernible on projectors (orthogonal: order the color key to match presentation of data categories)

For people who are colorblind. They are a large part of any audience.
posted by srboisvert at 9:16 AM on May 17


For data presentation, these are my go-to guys.
posted by Mental Wimp at 10:22 AM on May 17 [2 favorites]


I would put 'avoid 3D graphs' as 'avoid 3D graphs unless the 3rd dimension is used to present additional information that could not have been more simply presented in 2 dimensions' but I would propose that anyone who needs to be told not to use a 3D pie-chart isn't yet ready to use a 3rd dimension effectively.
posted by RobotHero at 10:41 AM on May 17 [1 favorite]


*Reads the first page, realizes this breaks every journal guideline ever, stops reading*.

Also: Labels on the graph? NO BAD. That way lines unclear placement and visual confusion!
posted by Canageek at 11:03 AM on May 17 [3 favorites]


« Older No one loves his bucket more than Charlie...  |  NOTEBOOK WITH DECORATIVE COVER... Newer »


This thread has been archived and is closed to new comments