Evidence based assessment/Poster Party
|Poster Party, hosted at University of North Carolina|
|Steps 1-2: Preparation phase|
|Steps 3-5: Prediction phase|
|Steps 6-9: Prescription phase|
|Steps 10-12: Process/progress/outcome phase|
Welcome to the Poster Party landing page! We will use this page to include a suggested syllabus for posters and non project-specific material.
The bulk of the work happens in between our weekly meetings, and there are several things that should happen in the coming weeks to get the most out of the party.
For the flipped format
I have pulled together a bunch of sample posters (all built in Powerpoint), and put them in DropBox. They are in two subfolders.
The “NCPA Student Posters” has a set of examples that actually were presented at NCPA over several years. They use a UNC template and colors. They look professional. They also reflect my sense of style, which is to reserve color for data, and use the ink to tell a story (heavily influenced by the work of Edward Tufte). So they look more plain or subtle than a lot you’ll see, and they are emphasizing the data.
The one labeled “Style” is a set of some greatest hits built by Guillermo Perez Algorta. They are PDFs because he used a different program to build them. They are visually much more engaging than the first set. There are things that could be tweaked about any of these – none is perfect. But what is exceptional about his work is how he is doing something visually interesting that supports the story of the data. The computer makes it easy to add clip art, faux 3D effects, and all kinds of junk. Adding junk just because we can actually distracts from the story and the data, and weakens the poster. I have seen thousands of posters in my life (quite literally). These examples from Guillermo are ones that I can close my eyes, see the poster in my mind’s eye, and remember what it was about – the visual image connects to the story. This is sadly rare. Don’t add junk. Challenge yourself to do something that supports the narrative, or don’t do anything – simple is fine.
Your action plan
- Review the examples (here’s a viewing link to the Dropbox folder)
- Pick one that you thinking is a good starting point from the “NCPA examples” folder, and do a save as. Give it a name like LastnameNCPA2017v1.
- Type over it – copy and paste your author list, your affiliation, your introduction, methods, results, discussion over the respective text boxes. (Pasting unformatted text will save you steps – it will keep the font and size from the example, rather than using 12 point Times or whatever was in the Word doc for the abstract). Voila! You have a rough draft of the poster. Seriously.
- Look for places to spruce up content. You don’t need to do much here. Putting too much text on the poster makes it less likely people will read. But sprinkling in a few of these can spice it up. Possibilities can include:
- Add (a bit) more detail to any section
- Add more citations
- Add more information from the results
- Add more future directions
- Look for places to spruce up style. Changing the starting template is a quick way to do this. I don’t recommend it for your first poster, because it can easily eat up hours. It’s better to experiment when you aren’t under a deadline.
The other advantages of sticking with one of the NCPA/UNC templates is branding – it reinforces that you are part of a team, and we get to benefit from UNC’s reputation. When you walk around, you will see that our median poster is of a different quality than the median of the rest of the conference; other people know that from previous years and when they see it’s a Carolina poster (Old Well or Ramses are great visual cues), they start with a different set of expectations. (I hope that App State has some posters at NCPA; they do an exceptional job of visual branding, and frankly, I think that their palette and design is cooler than Carolina’s. Love them for their style; love us for the quality of our science!)
- Look for pictures (not clipart) that support the story you are trying to tell (like the homeless person on Kay’s stigma poster, or any of the pictures in Guillermo’s selections)
- Look for figures from articles or websites that provide context for your results. This is a relatively new tactic for me. The gist is to use a picture to provide part of the introduction or discussion. The Gapminder grabs in Kay’s mood poster, or the circular image for the Big Five in Diane’s, are two examples. We always want to use these with attribution (plagiarizing on a poster is a spectacularly bad idea). You guys can now log into the UNC library web site and look at the e-copies of articles and snag the figures from the article, letting you paste them in and change size and still have them print beautifully. (When I was your age, the same result would have required a lot of experimentation with a photocopier to blow up the image, and then scissors and spray glue to add it to the poster. So no one bothered. Plus, dinosaurs didn’t like spray glue! ;-)).
ALWAYS add a citation to the source if you use this tactic!
- You could play with the font; but I mostly prefer to stick to classics like Arial/Helvetica/Calibri (sans serif) or Times (serif). Stuff that looks fancier tends to be harder to read, and it also distracts viewers into thinking, “Wow, that’s an unusual font!” instead of “Wow, that’s a cool question!” or “Wow, that’s a neat result!”
- Flipped part:
- Come with a copy of your poster ready to show – “Here’s what we’ve got: Any feedback?”
- Have a list of questions written out – what do you want feedback about? (Next level skill: You could embed these as comment bubbles on the draft)
- Make a list of any additional analyses you want or details from the results
- MOST IMPORTANT: Think about what would be the coolest figure to show the data for your main result. This is what I will be on tap to help you build.
- First step: What is the punchline of the poster? What is the one thing you want readers to remember?
- Second step: What was the statistical analysis? Use that to guide choice of graphic.
- t-test or ANOVA? Something that shows central tendency for groups is the graphic. Most of the world would use bar charts. Ho-hum. MECCA would use box & whisker plots (see Diane’s BFI poster) or even snazzier, beeswarms and boxplots (I think that we did this in Katherine’s poster, but you’ll have to check me). Why are these cooler? They show more data, and some let you pack even more information into the same space. Beeswarms show every data point, and you can use the size, color, and fill to push three extra variables into the plot. (HowAsian could be shade of color, sex could be solid fill/empty, and dot size…?). Plus, most people don’t know how to do it. It’s the same tactic as Guillermo’s really memorable pictures that supported the narrative; but here it is using the data visualization to show off in a way that supports the story. Style points for visualization!
- If the punchline is a correlation, then some variation of a scatterplot is the way to go. A scatterplot matrix is one snazzy option (see Chloe Paterson’s poster). A scatterplot where we play with dot size, shape, color, fill (forgot shape – we have a fourth dimension!) could let us cram 6 variables into one plot.
- If the punchline is a ROC, then the ROC plot is its own eye candy of science. SPSS produces adequate ROC plots. A lot of doubleclicking on them inside SPSS will let you customize them into something better. pROC kicks SPSS butt; trounces it, thrashes it, obliterates it… you get the message, but look at these examples and see that it is not hyperbole. No freakin’ contest. ROCR is another R package that would give a ton of neat options. This particular example has a bit too much going on for my taste, but no question that it is interesting and shows a lot of the data. Both the pROC and ROCR examples show the code they used to build the figure, which is key for us to be able to reverse engineer something similar with our data.
- Categorical variables present a challenge if they are the main event. Chi-squared is harder to make look visually interesting. I have been collecting ideas; none that I am completely sold on yet.
- For meta-analyses and translations projects, consider using an editable map. There are two ways of doing this. The easier is having an editable map object. The HCL poster has a world map that if you double click, it opens up and each country is its own object! This means that you can change the color for each country separately, which is exactly how we made that customized map. (Trivia fact: I happened to score that exact map from a slide deck Jules Angst shared with me. He’s the author of the HCL, along with being one of a handful of my heroes. Lots of serendipity there!). There is a programming way of accomplishing the same thing (or fancier), using chloropleth maps. These are GIS objects and libraries that we could link to R or SPSS and use for output. (Could be really neat to program into interactive web sites! Way beyond what we are going to do for NCPA, though – just planting a seed here).
- Bonus: Consider “breaking the fourth wall”: You could drop in comment bubbles or arrows and boxes on the printed version of the poster. It’s not done often, so it would be distinctive (like when Quentin Tarantino broke rules of the genre with Pulp Fiction). Could do it with editing and have it printed on the poster (Sabeen’s creativity Powerpoint deck, in the Examples folder, provides examples of this; though technically not a poster, the mechanics would be the same). I have thought about bringing sharpies to the poster session and inviting viewers to write comments on the poster (shading into interactive learning or performance art). It’s been a lot of fun the couple times I have done it, but I have been saving it for the last time that we show the poster (since it marks it up). Post-it notes could be a middle way.
- I will supply the infographic/visualization that you want. The more that you can tell me ahead of time what you want, the more prep work I can do ahead of time, and the faster this will go. Some of the stuff I have to use R instead of SPSS, so I’ll need to get the data into R ahead of time if anyone wants beeswarms, etc.
- After you have the infographic/visualization, play with the formatting so that it is all legible and looks good. Print a one page handout version. See if you can still read everything. Then decide if you actually *want* to read everything. Less is more with text.
- In the speaker notes, paste in the syntax that we used to build the figure and run the analyses. This guarantees that we can reproduce the figure, and it makes it much faster to borrow and recycle the analyses and figures (like we did with the Powerpoint template). You’ll see this in some but not all of the examples. This is a relatively recent innovation in the lab.
- Email around and get final feedback. NCPA uses a standard form for the judges; I have put a copy in the dropbox so you can see the exact rubric that they will be using.
- Send to print. I recommend PhD Posters, and we can get a group discount if we send them 10+ as a set. It will cost $$ (~$45 for a 48”x36” poster), but be cheaper than student stores or Kinkos. Save the receipt – those of you presenting again at a national meeting may be able to get reimbursed by the Office of Undergraduate Research (which will count as a grant if you get it – a third thing to list on your CV!)
- Practice presenting. Aim for 2 min or less. If people want to know more, let them ask. If you have multiple authors presenting, then everyone should practice doing the whole shtick and take turns presenting. Do NOT divide it up and have Eric present the intro, Guillermo present the Methods, Anna do the results…. It looks amateurish. You’ll see it happen at other posters. Don’t go out like that!
(13)BONUS: Consider recording your narration paired with the slide. Could do inside Powerpoint, or with Jing or other software. Vending machine bonus for whomever sends me a recorded version. Sushi lunch (or your preferred equivalent) to whomever puts it up on YouTube!
Always make sure to thank your funders (if you ever want to get money again!). For everyone in the ROC party this round, that means including this line: “(Grant Number: NIH R01 MH066647, PI: Eric A. Youngstrom)”
We are working on some MECCA branding and links (tinyurls, hashtags). These are going to be most important for the ROC and Wikipedia posters. If we have them ready soon enough, we could drop into the other posters, too.
Timeline and materials