08 September 2022

Box plot makeover

This is a figure in a journal.

Box plot with four boxes, each a different colour. X axis categories are "Unwashed control, washed control, unwashed experimental, washed experimental."

This is simple data that is much harder to interpret than it needs to be.

At a glance, it looks like it’s showing one variable with four groups. But it isn’t. To get that, you have to read each of the four labels on the horizontal axis very carefully.

Then you realize that the graph is showing two variables. It’s a 2⨉2 experimental design. One variable is environment and one variable is washing.

To make the relevant comparisons, I have to keep checking the axis label, because there is no other clue as to which box pairs with any box.

Significant differences are shown by three lines with asterisks above them. But the differences can be summarized in one statement: “Third from the left is higher than all the others.”

Here’s a quick and dirty graph makeover. As is so often the case, problems are solved by making things simpler.

Box plot with 4 groups. X axis: "Control" under two left boxes, "experimental" under two right boxes. Blue boxes labeled unwashed, purple boxes labeled washed. All boxes have an "a" above them except the third from the left.

First, I reduced the axis labels from four to two, one for each “environment.” Fewer labels also means larger labels, both of which make the axis easier to read. It more clearly shows what the adjacent boxes are comparing.

Second, I reduced the number of colours from four to two, one for each “wash” condition. This more clearly shows which non-adjacent boxes should be compared. (The colour and legend could be better. Remember, this is a quick makeover.)

Third, I swapped the lines for letters. The rule is, “groups with the same letter don’t differ.” This does lose a little information. The original has some comparisons with three asterisks, some with two, which usually indicates different p values. But that level of detail can be put in the text if it’s that important. It usually is not.

(Aside: The graphing program OriginLab has “paired comparisons” as a built in option.)

The design principles at play in this makeover? 

First, be cautious of templates. The original graph looks like someone just used the default settings in a graphing program. (R studio, maybe ggplot2?) 

Second, simplify. 

Third, make related things similar. Usually, I say, “Keep related things together”, referring to similar positions in space. That is in play here with the axis labels. But related things are also shown by other similarities: colour, shape, and so on.

External links

Original Twitter thread

01 September 2022

One class of pharmacy students preferred online poster sessions

We spent two years mostly having conferences online. Many people see advantages to that format and want conferences to keep online options. But would online conferences be just a better-than-nothing kludge for people who would normally be left out? A roughly equal experience to traditional conferences? Or could it be that the online experience is actually superior to face-to-face presentations?

These questions matter more for poster sessions than oral presentations. Oral presentations online are a reasonably solved problems. TED talks have shown for years that a good recorded presentation can have impact.

A new paper compared students’ preference for online and face-to-face poster sessions. Axon and Whaley (2022) found people liked online poster sessions more.

If the graph below, the blue on the left shows people who preferred online, the tan in the middle shows people who preferred face-to face, and the purple on the right is no difference. Click to enlarge!

Student pharmacists’ preferences for a virtual versus in-person research poster session. Figure legend: Acquire development skills: I would acquire more poster development skills if the research poster session is. Acquire presentation skills: I would acquire more poster presentation skills if the research poster session is. Acquire participating skills: I would acquire more skills participating in the poster session if the research poster session is. Enjoy developing: I would most enjoy developing a research poster if the research poster session is. Enjoy presenting: I would most enjoy presenting a research poster if the research poster session is. Enjoy participating: I would most enjoy participating in a research poster session if the research poster session is. Able to develop: I would be most able to develop a research poster if the research poster session is. Able to present: I would be most able present a research poster if the research poster session is. Able to participate: I would be most able to participate in a research poster session if the research poster session is. Effectively communicate findings: The most effective way to communicate findings from a research project is. Effectively evaluate findings: The most effective way to critically evaluate findings from a research project is. Effectively participate: The most effective way to help ensure everyone has the opportunity to participate in the research project poster session is. Overall preference: Overall, if I had to choose one approach for developing a research poster, presenting a research poster, and participating in a research poster session, I would choose.

More people preferred online over face-to-face in every category except one. People thought they had more ability to acquire presentation skills in a face-to-face setting (second bar from top). But even there, the advantage is not huge.

But there are many issues that need to be pointed out here!

This paper is the result of surveying one class of pharmacy graduate students. And I don’t mean, “a single course that has been taught multiple times,” I mean one group of students in one semester. The sample size is 63.

Critically, those students did not have experience with both formats in their class. They had to give their posters for their class in an online format. This is a “How would you know you liked it if you never tried it?” situation. Maybe some students had experience with face-to-face poster sessions before they took this class, but the survey didn’t ask.

It’s unclear what the online poster session for these students was like. This is the whole description:

The virtual format required students to record a presentation of their poster and provide a link for their peers to review and comment on during the research poster session.

Online poster presentation platforms vary wildly. Some are good, some are not so good. Maybe the team teaching this course managed their online poster session extremely well, and that was reflected in students’ responses. Opinions don’t exist in a vacuum.

It’s great that people are open to the possibilities of online poster sessions. But this paper tells us little about whether people who have experience with both formats prefer. And that seems like an important question to ask.

References

Axon DR, Whaley M. 2022. Student pharmacists' perspectives of in-person versus virtual research poster presentations. Pharmacy 10(5): 104. https://www.mdpi.com/2226-4787/10/5/104

31 August 2022

“Design a poster” interview with Tullio Rossi

Friends of the blog Tullio Rossi and Echo Rivera (who penned the intro to Better Posters) talk posters!

This should have been in the last month’s link round-up. Whoops. But I don’t want to wait several weeks to get it into the next one, so it gets a stand alone post!

25 August 2022

Link roundup for August 2022

There’s a lot of device on making graphs, but sometimes research on whether the advice works is harder to come by.

This research paper on graph design tries to assess whether “decluttering” (Edward Tufte’s “data to ink ratio”) and “focus,” which is annotation of graphs. I have to say “try” because the sample size for the experiment is just 24 people, which seems small for a topic like this.

Dot plot showing ratings for the three visualization design on a scale from 1 to 5 on aesthetics, clarity, professionalism, and trustworthiness. Focused graphics rates highest in all four categories.

“Focusing” on a graph was very effective in helping people remember. 

Removing clutter had a smaller effect, but people generally liked those graphs a little better.

Ajani K, Lee E, Xiong C, Nussbaumer Knaflic C, Kemper W, & Franconeri S. Declutter and focus: Empirically evaluating design guidelines for effective data communication. IEEE Transactions on Visualization and Computer Graphics: in press. https://doi.org/10.1109/TVCG.2021.3068337

This is a plain English summary of the paper

• • • • •

People like to attack graphs with a lot of noise in this. Stuart Ritchie reminds us, “Noise is why we use statistics

 And that’s it for this month!

18 August 2022

Review: Storytelling with Data

Cover to book "Storytelling With Data"

Some scientists might pass over Storytelling With Data (2015) because they have an incredible aversion to the word “storytelling.” Author Cole Nussbaumer Knaflic might drive off some people when she writes things like, “The magic of story.” It can sound a little hippy dippy trippy sometimes.

If “story” and “storytelling” freak you out, let me make a pitch for this book and give you another way to look at it.

Most of the book is about annotated data visualization. Fully seven of the book’s ten chapters are devoted to picking apart and recreating graphs. 

For example, Knaflic takes a mundane bar graph like this:

Bar graph of annual retail prices per year.

And shows the process to get to this:

Annotated line graph of annual retail prices per year.

Same data set. But one has a much clearer focus on what information is critical to the viewer.

I sometimes do similar things with graphs here on the blog, and some of the principles will be familiar to readers of the Better Posters book. But this book drills down into making a single graph in a much deeper way, with many more examples.

How to take a single graph from an Excel default and into something more targeted and informative has been on my mind a lot lately, and this book is the best I have found on the subject so far. Recommended.

Related posts

Same graph, different narratives

External links 

Storytelling with Data website

16 August 2022

Conference presentation accessibility

SurveyMonkey logo
I am helping with a project on conference accessibility. I was brought on board because of my interest in posters, but this research is about oral presentations, too.

To get us started on data collection, we have a short (~10 minute) survey on how presentation format impacts people with accessibility needs at academic conferences.

You can take the conference presentation accessibility survey here: https://www.surveymonkey.com/r/2DNJ2NX

Or you can use this easier to remember shortened link: https://bit.ly/ConfTalkAccess (Short for “Conference Talk Accessibility.”)
 
Or you can scan this QR code:

QR code leading to conference accessibility survey

This survey will be available through 6 September 2022.
 
Please take the survey. Please share the survey with others who might go to conferences. Thank you for your support! 😁

11 August 2022

Expectations and visual noise: Lessons from washing machines

“How do I turn this on?”

I was visiting, trying to use the washing machine, and was confronted with this.

Washing machine button with instructions in fine print at the bottom.

When you want to turn on something with a knob, you look for something around the dial that “Start” or “On,” right? Nope.

But maybe this was more a button, and when you want to turn on something with a button, you press it, right? Nope.

I couldn’t get the washer to start. I literally gave up on fiddling with the machine. Instead, I:

  1. Opened my phone
  2. Typed in the maker and model of the washing machine
  3. Found an owner’s manual online
  4. Downloaded the PDF of the manual
  5. Skimmed through the first few pages until I found the instructions to turn on the machine.

No joke. And I did all this even though the answer was staring me in the face.

If you look at the button above, you will see the instructions to turn it on. And there’s the first problem.

“Pull start.”

You mean, the complete 180° opposite of the usual thing that people do to turn on machines? Maybe something that runs so counter the norm is important!

Lesson #1: Warn people – loudly – if something works differently than the way most other things work, warn people. Not in a whisper.

But even putting that aside, why could I not see the instructions that were helpfully right by the button? There’s the second problem. The instructions are in fine print, surrounded by a lot of other information about four other wash cycles, at the bottom.

Lesson #2: More information can make a task harder, not easier.

This washer button reminded me of so many conference posters I have seen over the years. Things are not in places readers expect them to be. There is so much information and so little emphasis that readers are struggling to find what they want.

My case was unusual, because I was using this machine for the first time. You do expect the same person to be using a home appliance over and over again. The “fine print” instructions are normally only going to serve as a reminder.

But I had no problem turning on the dryer.

Dryer buttom with ““Push to start” in large bold letters.

Less information, and what is present gets some emphasis using size, bolding, and colour.

This is more like what a conference poster should aspire to be.

Drake turning away from hard to use washer button and approving of easy to use dryer button.

A shorter version of this originally appeared as a Twitter thread.