We interviewed a few current and former officers of the Statistical Graphics Section to find out why statistical graphics are an important component of the profession. Here’s what they said.
Mike Jadoo
Economist, Bureau of Labor Statistics
Why did you choose to be a statistician/data scientist?
I enjoy problemsolving and statistical analysis, which led me to a career in data science. With statistics, you get to do projects and research that have direct applications in the world. You can help solve a problem directly.
Why did you join the Statistical Graphics Section?
To network and gain a better understanding of the proper use of statistical graphics in statistics and data science.
What have you gained from being a member of the section?
A better understanding about how and why to use data visualizations in reports and dashboards.
List three of your favorite data visualization tools and tell us why you like them.
Horizontal bar charts—they show lots of categories in one chart that is easy to understand.
What advice would you give someone starting their journey in statistical graphics?
- Learn the basics. Start by familiarizing yourself with the basic concepts of statistics and data visualization.
- Choose the right tools. There are many tools available for statistical graphics, such as R and Python (with libraries like Matplotlib, Seaborn, and Plotly), and software like Tableau and Excel. Choose one that suits your needs and preferences and start practicing with it.
- Practice, practice, practice with real data. Theory is important, but practical experience is crucial. Start working with real data sets to apply what you’ve learned.
- Focus on interpretation. Don’t just create visualizations. Focus on interpreting the data accurately and effectively communicating insights. Ask yourself what story the data is telling and how your visualization can best convey that story to others.
- Learn from others. Read books, articles, and blog posts and follow practitioners on social media.
- Stay curious and keep learning.
How do you see statistical graphics evolving in the next decade?
Statistical graphics will be integrated with machine learning techniques to provide deeper insights into complex data sets. This may include the use of algorithms to generate visualizations tailored to specific data sets or user preferences.
What three favorite books do you recommend to others who have an interest in statistical graphics?
- ggplot2: Elegant Graphics for Data Analysis by Hadley Wickham. If you’re interested in using R for data visualization, this book is indispensable. Wickham, the creator of the ggplot2 package, provides comprehensive guidance on creating elegant and sophisticated graphics using ggplot2.
- Information Visualization: Perception for Design by Colin Ware. Ware explores the cognitive principles underlying effective information visualization. This book provides valuable insights into how visual perception influences the design and interpretation of statistical graphics.
- Storytelling with Data: A Data Visualization Guide for Business Professionals by Cole Nussbaumer Knaflic. This book focuses on the storytelling aspect of data visualization, offering practical tips and techniques for creating compelling narratives with data.
Lucy D’Agostino McGowan
Assistant Professor, Department of Statistical Sciences, Wake Forest University
Why did you choose to be a statistician/data scientist?
My father and grandfather are both statisticians, so I grew up knowing what a great field this is. However, it wasn’t until I completed the Boston University Summer Institute for Research Education in Biostatistics that I really began my journey into biostatistics. I love being a statistician because it gives me the opportunity to help solve different problems across different domains.
Why did you join the Statistical Graphics Section?
I spend a lot of time thinking about how to best communicate complex statistical information to a variety of audiences. Being able to communicate via accurate, useful, and intuitive data visualizations is a key component of this.
What have you gained from being a member of the section?
Through this section, I have met so many statisticians I deeply admire who are similarly interested in finding the best way to communicate complex statistical information.
List three of your favorite data visualization tools and tell us why you like them.
When the intended audience is me, the analyst, the classic scatterplot is always a go-to. Any visualization that gets all the individual data points in front of my eyes is key for making sure I have a good understanding of several pieces of information like the range and spread of the data, any potential nonlinearity, etc. In terms of software, the R package ggplot2 is my go-to for building static visualizations. It offers a high level of customization, and I love the grammar of graphics style, which makes it straightforward to teach and use. For interactive visualizations, I really like Highcharts.
What advice would you give someone starting their journey in statistical graphics?
Find a community to support your journey! There are so many amazing statisticians who are willing to help incoming folks learn and grow in our field. Finding a community can really accelerate your progress (and it’s fun!).
How do you see statistical graphics evolving in the next decade?
I hope we continue to learn from the audiences consuming the information we generate. The audience is a key component to consider when deciding how to communicate information, and as the audience changes, so must we.
What three favorite books do you recommend to others who have an interest in statistical graphics?
There are so many great books, and no book is perfect, but a few that have influenced my work include the following:
- The Art of Insight: How Great Visualization Designers Think by Alberto Cairo
- Fundamentals of Data Visualization: A Primer on Making Informative and Compelling Figures by Claus Wilke
- The Visual Display of Quantitative Information by Edward Tufte
Edward Mulrow
Senior Vice President and Director, Statistics and Data Science, NORC at the University of Chicago
Why did you choose to be a statistician/data scientist?
From high school through my early graduate school days, I had an interest in applied probability. As I learned more, I became interested in statistics and data analysis. Working in the government and as a federal statistical agency contractor provided me with an appreciation for quality data and proper analysis of that data.
Why did you join the Statistical Graphics Section?
I have always found graphics interesting, whether it was plotting a function in algebra class or visualizing data in statistics class. Because of that interest, I was drawn to JSM sessions related to graphics. I learned those sessions were sponsored by the Statistical Graphics Section, and I joined the section because others in the section had similar interests to mine.
What have you gained from being a member of the section?
The section provided me with an opportunity to be involved in the American Statistical Association and helped foster a great network of colleagues. Naomi Robbins, Rich Heiberger, Di Cook, and Heike Hoffman are a few of the people I have met through the section with whom I have had the good fortune to collaborate on projects.
List three of your favorite data visualization tools and tell us why you like them.
Systat was the first tool I regularly used for data visualization. I don’t use it anymore, but I think it created good graphics. Systat was created by Leland Wilkinson, who was a big proponent of statistical graphics and often gave good JSM Statistical Graphics Section presentations. The Systat manual not only laid out how a user could create a good graphic but also explained the purpose of the graphic along with literature references.
Hadley Wickham’s ggplot2 R package is my main go-to visualization tool. The “gg” stands for The Grammar of Graphs, which is a book by Wilkinson that provides an object-oriented way of thinking about statistical graphics. I find ggplot2 easy to use and intuitive. Once I have a basic graphic designed, it is easy to add or remove features so the graphic suits my needs.
My third favorite tool is my colleagues. Not only can I draw upon them to create graphics using tools for which I have little expertise, but they are a wonderful group for feedback on my ideas. My favorite graphics are ones created with a team of colleagues who all interjected their thoughts into the process of creating an effective graphic.
What advice would you give someone starting their journey in statistical graphics?
Remember an effective graphic must be understandable to those in the intended audience. Creating effective data graphics can become technical, and some visualizations require complicated programs and techniques. Don’t get too engrossed in the programing without considering how viewers will perceive the visual.
Critique your work and ask colleagues for their opinions on the effectiveness of your graphics.
My colleague Nola du Toit and I have a paper titled “Visual Communication of Data: It Is Not a Programming Problem, It Is Viewer Perception” in Computational Statistics in Data Science. Our advice is that creating effective graphics is not just about the technical tools for creating data visualization. It is also about the conceptual and perceptual tools you use to develop and evaluate your graphics.
How do you see statistical graphics evolving in the next decade?
There are a lot of hot topics on the horizon, including access to more data, improved tools for creating visualizations, data storytelling, and many more exciting concepts.
I think it would be good to take a step back and reconsider the makeup of graphics and the expert advice on creating effective graphics.
My NORC colleagues Kiegan Rice and Nola du Toit, along with Heike Hoffman from the University of Nebraska-Lincoln, are reconsidering the question of what viewers see when they view a chart. They plan to build on the foundations of Cleveland and McGill’s work in graphical perception by employing a large, nationally representative, probability-based panel of survey respondents to test visual perception. Their approach affords the ability to dive into response patterns among demographics across perceptual tasks and identify significant differences among them.
Past studies that used nonrepresentative or crowd-sourced samples may have missed how viewers with different demographics, such as lower education or lower income, perceive statistical graphics. If so, we may need to come up with a revised playbook on the makeup of a good statistical graphic.
What three favorite books do you recommend to others who have an interest in statistical graphics?
- Visualizing Data by William S. Cleveland. After seeing so many references to Cleveland’s visualization work, I decided to attend a visualization workshop present by him. I thoroughly enjoyed the workshop and could not wait to get his book. I read it cover to cover in a short amount of time. While there have been many advances in visualization since this book was published, the concepts laid out in the book are still relevant today.
- Creating More Effective Graphics by Naomi B. Robbins. This book is a great reference. Robbins provides step-by-step instructions for creating effective graphics. I have often heard her joke that her book is “Cleveland for dummies.” While the book is written in a way that reaches a diverse audience, even the most knowledgeable visualization expert will find the book useful.
- Visualizing Data Patterns with Micromaps by Daniel B. Carr and Linda Williams Pickle. This book introduces a concept for including geography in a high-dimensional data visualization. Choropleth maps are good but don’t handle multiple variables well. The use of micromaps is one way to overcome this problem. The book describes why micromaps are effective and provides many examples of different ways they can be used.
Joyce Robbins
Lecturer in Discipline, Department of Statistics, Columbia University
Why did you choose to be a statistician/data scientist?
I’ve always had a broad range of interests with a special pull toward math and computers. Statistics and data science have a role to play in just about everything in modern life, so it’s a natural fit.
Why did you join the Statistical Graphics Section?
I joined to learn about current work in the field and meet people with similar interests. In addition, given the size of the JSM conferences, being a member of a section appealed to me since it provides a sort of “home base”—a smaller group with its own events, sessions, etc.
What have you gained from being a member of the section?
One of the most rewarding aspects of being a member of the section has been organizing sessions on new research directions and connecting people who work on similar topics. One of these sessions addressed the use of JavaScript in statistical graphics with speakers Karl Broman, Carson Sievert, Ramnath Vaidyanathan, and Joy Yang. At another, Jason Cory Brunson, Ursula Laa, and Hengrui Luo presented their latest developments in graphing multivariate data. Organizing a session gives you the power to custom build the panel you would most want to attend at a conference. It’s rewarding and a great way for newcomers to the section to get involved.
List three of your favorite data visualization tools and tell us why you like them.
Number one is R/ggplot2. This is my go-to tool for statistical analysis and static visualizations. I could talk endlessly about the power of R, but I’d be preaching to the choir with this audience.
Number two is D3. I love the ability to make unique interactive visualizations, particularly for teaching.
Number three is pencil and paper. Perhaps I should have listed that first because it is so important to think about how you can best represent your data without the limitations of your coding skills or the constraints of your go-to graphics software. It’s worth remembering in this context that John Tukey’s Exploratory Data Analysis, the seminal work in statistical graphics, is based largely on paper and pencil methods.
As different as these tools are, they are all incredibly flexible. None come with a list of chart types but let the user design what works best for their use case. D3 and ggplot2 hit the right balance between high-level and low-level coding, so one can design custom graphics without getting too into the nitty-gritty details (usually).
What advice would you give someone starting their journey in statistical graphics?
Statistical graphics is a field with theory and principles that should always inform one’s work. Sometimes, students measure their skill set in terms of the number of graphics tools they know, which is misguided. Keep an open mind and look for opportunities to learn both within the statistical community and outside it. Pay attention to the people who are doing work that interests you and those who are writing the packages you find useful. Follow them on social media and look for opportunities to hear them speak at conferences, webinars, and meetups.
How do you see statistical graphics evolving in the next decade?
Statistical graphics is already expanding from data exploration to model exploration. For example, it is critical in the developing field of interpretable machine learning. I think this trend will continue. In addition, it almost goes without saying that AI will play a bigger role, as it will in so many technical areas.
What three favorite books do you recommend to others who have an interest in statistical graphics?
For those with little experience with statistical graphics, including nonstatisticians, I recommend Creating More Effective Graphs by Naomi B. Robbins, which provides practical tips for improving graphs mainly based on [William] Cleveland’s work on visual perception. (Full disclosure: I am Robbins’ daughter). For those with some knowledge of R who want a comprehensive introduction to the field, I recommend Graphical Data Analysis with R by Antony Unwin. Finally, for a more in-depth look at visualization techniques for those with a basic foundation in statistics, I recommend Visualizing Data by William S. Cleveland.
Emily Robinson
Assistant Professor, Department of Statistics, California Polytechnic State
Why did you choose to be a statistician/data scientist?
Growing up, I enjoyed mathematics and attended Winona State University in Minnesota for secondary math education. When I started college, I knew little about statistical analysis and data science. I approached statistics from a black-and-white perspective. Thanks to the professors there, I gained a new perspective on analytical thinking and problem-solving as I realized the world was variable and subject to change.
Statisticians are challenged to make sense of constantly changing information and communicate this to the public. This was a challenge I wanted to accept. I added a statistics major, did a few summer internships in statistical industries, and ultimately earned my PhD in statistics at the University of Nebraska-Lincoln. I am now entering my third year as an assistant professor of statistics at California Polytechnic State University, where I enjoy teaching, mentoring students, and collaborating on research projects.
Why did you join the Statistical Graphics Section?
I began researching the perception of statistical graphics after my PhD adviser, Susan VanderPlas, printed out a paper on visual inference and graphical testing. I enjoyed the studies and read all the related papers over the next week. Subsequently, I joined the Statistical Graphics Section to connect with others doing similar work.
What have you gained from being a member of the section?
As part of the Statistical Graphics Section, I have built relationships with researchers and individuals interested in improving statistical graphics. It is a tight-knit and welcoming community. I have been invited to give conference presentations, lead workshops, and take on leadership roles through the section.
List three of your favorite data visualization tools and tell us why you like them.
- d3.js. I have recently enjoyed learning D3 as a tool for creating interactive graphics. This is used to create graphics in the data journalism sector, and I am intrigued by using it to understand user estimation processes.
- Dear Data. While this is not necessarily a tool, this book follows the journey of Giorgia Lupi and Stefanie Posavecas as they create a pen-pal relationship by sending postcards of hand-drawn data visualizations about their lives that week. I love using this tool in classes to get students thinking outside the box when displaying and communicating data.
- ggplot2. This wouldn’t be complete without a shout-out to ggplot2. I enjoy how this tool follows the grammar of graphics and layers each aspect of the graphic for easy adjustment between types of charts or separating groups with color or facets. Once you learn the language, you have a lot of freedom in creating high-quality graphics or quickly doing exploratory data analysis.
What advice would you give someone starting their journey in statistical graphics?
Join the TidyTuesday community! Each week, TidyTuesday posts a new data set that is often related to society at that time. Participating in TidyTuesday helps you practice your skills and allows you to see how others approach the same data set. Grab the data set, your favorite tool, and create!
How do you see statistical graphics evolving in the next decade?
In recent years, we have been able to generate graphics quickly, which increases the need for guidelines in creating effective graphics. Moving forward, I think we will continue improving graphics’ readability, especially around accessibility (e.g., screen readers, 3D printers, etc.). Additionally, I can see us heading toward more interactive graphics, giving graphic users the autonomy to explore.
What three favorite books do you recommend to others who have an interest in statistical graphics?
- Show Me the Numbers: Designing Tables and Graphs to Enlighten by Stephen Few
- Better Data Visualizations: A Guide for Scholars, Researchers, and Wonks by Jonathan Schwabish
- Fullstack D3 and Data Visualization: Build Beautiful Data Visualizations with D3 by Amelia Wattenberger
Susan Vanderplas
Assistant Professor, University of Nebraska, Lincoln
Why did you choose to be a statistician/data scientist?
As an undergraduate, I wanted to be anything other than a statistician. My dad is a statistician (he’s now retired) and, obviously, the stuff he did all day was boring.
After trying my hand at cognitive psychology as an undergrad and bioinformatics as a grad student, I realized I liked working with data—I just didn’t want to be confined to linear models and significance testing. Luckily, there’s more to being a statistician than ANOVA and linear regression (something I definitely didn’t realize as a teenager).
I ended up combining my interest in cognitive psychology with my interest in data and primarily work on problems related to the intersection of human perception, algorithms, and data visualization.
Why did you join the Statistical Graphics Section?
It’s a fun place to be! I think I officially joined after I was recruited to represent the section in the Council of Sections but, as a student, I always went to the Graphics/Computing mixer at JSM because it was fun. It’s still one of my favorite parts of JSM because I get to see all the people in my field in one place and hang out.
More generally, graphics are an important part of how statisticians communicate with the rest of the world and it’s good to interact with people interested in improving how statisticians use graphics.
What have you gained from being a member of the section?
A chance to network and make friends across the space of graphics and visualization. I’ve also had opportunities to develop workshops, serve on award committees, and put together sessions at JSM and the Symposium on Data Science and Statistics relating to graphics and data visualization.
List three of your favorite data visualization tools and tell us why you like them.
- ggplot2 is my main tool for creating charts. It’s lovely because of the clear implementation of the grammar of graphics, but it also has a ton of useful packages that extend its functionality.
- My next tool is my tablet and stylus or a set of felt pens and paper—sketching is probably the next most important tool I use regularly.
- Finally, shiny is a great interface for creating interactive displays. I use it for everything from data collection at the beginning of a project to data and model exploration to presenting results intended for others’ consumption. As an interactive graphics tool, it’s definitely limited in scope, but the features it has are extremely useful for many common tasks.
What advice would you give someone starting their journey in statistical graphics?
You’ll almost always end up needing to learn multiple tools and/or programming languages. As you pick up new tricks, though, it’s useful to think about how something is implemented in R and ggplot2 versus python and seaborn, for example, and use those comparisons to build a concrete understanding of both the tool and wider theoretical space. The grammar of graphics looks different in ggplot2 than seaborn objects or VegaLite, and recognizing those differences can help you pick the right tool for a job, but they can also help when you set out to design a better tool.
How do you see statistical graphics evolving in the next decade?
One area of research I expect to take off is in nongraphical representations of data. Data sonifications, for instance, have to be interactive because sound is instantaneous, so designing a system/grammar for creating data sonifications not only makes data more accessible to people with visual impairments, it also stretches concepts we take for granted about visual representations.
We also don’t know as much about perception of sound, touch, etc., so there is a lot of work to be done on how to leverage the strengths of other senses to communicate data-related findings effectively. I love the idea of being able to add additional information to a visual display through sonification, as well as the potential to leverage touch using 3D printed data representations.
I also hope we get better tools for interactive and linked charts that are as well supported as ggplot2 and matplotlib. Having the ability to highlight and follow points through different visualizations generated on demand and linked, as you could do with tools like ggobi, is something currently missing from the space of tools available for modern operating systems.
The tech nerd in me would also love to see an implementation like ggobi that works with headset-based tools—so you can generate plots and move them around a room space—with each plot linked as specified. I don’t have those kinds of programming skills, and it remains to be seen whether VR headsets will become standard tools for offices and video gaming, but I’d love to see something like that in the future.
What three favorite books do you recommend to others who have an interest in statistical graphics?
- Visualization Analysis and Design by Tarama Munzner. This is a great introductory textbook that covers the perceptual concepts important for creating many types of good visualizations and approaches to testing and validating visualizations.
- The Statistical Atlas of the United States. There are atlases for the 1870, 1880, and 1890 US Census. The atlases contain a visual exploration of the United States during westward expansion and come from a time before graphical conventions were standardized. Not everything they tried worked (I’ve written a paper about framed spine plots, for instance, which did not work very well), but the graphics are absolutely fascinating. It’s amazing to think about the processes they had to use to generate the color images—adding wax to stone and scraping it off precisely for each layer of the three to four colors of ink required to create a full color chart. I’m quite grateful modern tools don’t require that level of patience or artistic precision.
- Getting (More out of) Graphics: Practice and Principles of Data Visualization by Antony Unwin. This book has a lot of great examples of good graphics (and some fascinating data sets) and great discussions of the pros and cons of different design decisions and the insights that can be gained from different representations of the same data.
- Thing Explainer by Randall Munroe (a bonus book not specifically about statistical graphics). This isn’t a statistical graphics book at all. It’s written by the guy responsible for xkcd. In the book, he uses illustrations and the thousand most commonly used words in English to explain complex subjects such as how nuclear reactors work and the electromagnetic spectrum. What I love about this book (I keep a copy on my desk.) is it’s a great example of how visual illustrations and simple text can be combined to communicate technical information without confusing people.
Emily Zabor
Biostatistician, Cleveland Clinic
Why did you choose to be a statistician/data scientist?
I became a biostatistician because I wanted to use my strengths in mathematics and critical thinking to address real-world research questions. More practically, I knew it was a field with many job opportunities across a wide variety of applications, so I thought it would be a good long-term career in terms of job options and ability to find a position that would interest me no matter where I wanted to live.
Why did you join the Statistical Graphics Section?
I strongly believe in the importance of graphics to convey information effectively, so I joined the Statistical Graphics Section to connect with others with similar interests and to learn about cutting-edge developments in statistical graphics. For me, graphics are not secondary to the main results of an analysis but are a critical piece of the analysis itself to understand the data and convey the results.
What have you gained from being a member of the section?
I have met many other statisticians and data scientists with expertise in diverse areas such as color theory and accessibility and have been exposed to many graphical topics I had not previously considered. As a result, I am better able to create graphics that are not only more visually appealing, more consistent, and more informative but also more accessible to a broad audience.
List three of your favorite data visualization tools and tell us why you like them.
I do all my data visualization in R and love using the ggplot2 package for its layered syntax and customizability, the plotly package for creating interactive visualization, and the viridis package for its wonderful color palettes.
What advice would you give someone starting their journey in statistical graphics?
I would advise someone to elevate graphics from something done as an afterthought to a critical part of every data analysis and considered up front. Think carefully about what question you are trying to answer and how a graphic can help address it. I often write the pieces of information I want to convey and then think about how to map the various parts to shapes, colors, dimensions, axes, etc. I usually make many changes to my plans before I actually begin making graphics.
How do you see statistical graphics evolving in the next decade?
I think interactive graphics will continue to be used more frequently to convey more levels of detail than what can be achieved with static graphics. I also think statistical software to standardize and streamline the most common graphics will continue to improve so even the simplest graphics will become more readable and visually appealing with little customization.
What three favorite books do you recommend to others who have an interest in statistical graphics?
- Storytelling with Data: A Data Visualization Guide for Business Professionals by Cole Nussbaumer Knaflic
- Resonate: Present Visual Stories That Transform Audiences by Nancy Duarte
- Beyoncégraphica: A Graphic Biography of Beyoncé by Chris Roberts (just because it’s fun)
Leave a Reply