Welcome to my public portfolio for the Fall 2019 Telling Stories with Data course!
I first became interested in the social applications of data science through my engagement with the Effective Altruism and Data Science for Social Good communities, which seek to apply reason and evidence to determine the most effective ways to benefit others. My interactions with these communities opened my eyes to the depth of need that exists within the public sector both for more high-quality data as well as for the knowledge and skills to put that data to effective use.
This realization is what ultimately led me to Carnegie Mellon, and to Heinz College in particular. As the country’s top-rated program in analytics education with a mission of training students to “turn raw data into solutions for society’s most pressing problems,” the MSPPM program at Carnegie Mellon offered the perfect opportunity to develop the technical competencies I was seeking as well as an ideal environment to practice those skills in a real-world setting—namely, the City of Pittsburgh and its diverse network of organizational partners.
My most recent projects have focused on public interest technologies that help to increase the range and accessibility of social services worldwide, especially as a means of increasing society's resilience to global threats like climate change.
View my resume [here](https://jaxgoodlabs.github.io/patrick-campbell-portfolio/Patrick%20Campbell%20Resume.pdf).
In this course, I hope to learn how to tell more clear and compelling stories with data and make any audience feel like they can immediately engage with the stories I'm telling. I also look forward to the opportunity to learn the software that is setting the new standard of how data should be visualized in professional settings and to build a portfolio through which I can display these new skills to prospective employers.
Below are some examples of my work.
For this exercise, I chose to redesign a chart depicting economic inequality in the US, published by Atkinson, Hasell, Morelli, and Roser (2017) on the website “The Chartbook of Economic Inequality.”
I chose this visualization because I was interested in the topic of economic inequality and familiar with the narrative in general (i.e., that inequality in the US is on the rise), but not particularly clear on its finer details. When I found this graph and saw all the technical jargon it contained, I saw an opportunity to test the limits of what effective data visualization techniques could accomplish. I was confident that I could clean the chart up aesthetically, but I was less confident of how accessible I could make its content given the technical nature of the underlying data.
My critique process involved two steps. First, I evaluated the graph in its current presentation and took detailed notes on what I felt its main strengths and weaknesses were. It was clear off the bat that the graph's target audience were not laymen, but rather economists and audiences with high economic literacy. Even with that knowledge, however, I thought the graph was overly reliant on text descriptions. So much text seemed a bit unnecessary for those whom I imagined to be the target audience, and at the same time insufficient to produce understanding among everyday readers. I thought there may be a way to strike a middle ground by pairing a more common language description of each trendline with a more in-depth, but still condensed technical description of the metrics being used.
To test this assumption, I moved on part 2 of my critique process, which was to build a wireframe prototype of my redesign. Even after reducing and reorganizing the text, however, I was still unsatisfied. The graph was trying to accomplish too much at once. I thought about different ways to trim it down without losing too much of its meaning. I was too unfamiliar with the particular metrics to be very judicious here, but there were lots of other reasons to eliminate the earnings dispersion data - most notably, the fact that it was measured on a different scale than all the other metrics. This disparity in scale was creating a lot of confusion for what I felt to be not much payoff, so I decided to eliminate it. When I did, the meaning of the graph came into much sharper focus.
With every new iteration of my redeisgn, I immediately shared it with others for feedback. Much of the early feedback I received confirmed my own opinions - "the abrupt change of scale between earnings dispersion and all the other metrics is confusing," "this graph is not meant for normal people," "the color coding and text descriptions are helpful for understanding what these trends mean, but still aren't very clear," etc. One of my first revisions was to remove the key at the top in an effort to eliminate redunancies and reduce the amount of text (see example above), but one reviewer felt that change made the graph more difficult to understand. In my final revision, I reintroduced the legend as headers for the more detailed descriptions on the right margin, which received much more positive feedback. My final changes were to thicken the trendlines to mask the individual data points and add a more descriptive subtitle to better drive the graph's core message home, yielding a much cleaner, more accessible visualization of economic inequality in the US.
...And finally, the finished product created with Tableau (below). Because this was my first time working with Tableau, there were a few features from the prototype that I wasn't able to reproduce in the final version - namely, the interpretive arrows on the left margin and the expanded legend. Portions of the data were also clipped to preserve readability.
See my final project here.
Check out some of my smaller projects and experiments here.