Based off dataset: Access to food by US county
I chose this data set because I wanted to learn more about poverty levels around my county and in other counties. As I explored the datasets, I saw that this dataset explored households that did not have a vehicle. I personally do not own a vehicle and I thought it was something I could relate to and understand where people really struggle to get access to a supermarket.
I did data cleaning at different steps and phases within my project based on my changes from feedback and visualization ideas. The first step that I took was deleting and getting rid of all states except for Wisconsin. After, I eliminated counties.
Starting off, I immediately knew that I wanted to compare the total population to some sort of variable with housing access and vehicle access.
Above (figure 1) is my decision matrix that helped me narrow my ideas. During this phase I spent time working with the data dictionary (figure 2). The first idea that I had was to create a visualization of each state as a whole and compared all the data variables (see figure 2 for more information on the variables), however this felt out of scope, and I had to reduce the overall amount of data that I would be using. I needed to remember to keep my visualizations simple and my goal was not to display all the data at once. I started to develop a focus and really liked one idea overall; “housing units without a vehicle present.” I scanned through my data at this point and found that there were prevalent amounts of data in the houses with no vehicle that were also 1 mile and ½ Mile from a supermarket. The reason for this decision was because the other variables "10 miles or more" and "20 miles or more" often had values that were very low or zero.
Above (figure 2) is my data dictionary. The focus is the highlighted section in red. As noted, I chose to not use this data in my visualization as these variables overcomplicated the visualization that I wanted to create. My goal at this point for the visualization was: “Number of people without vehicle access that are 1/2 mile and 1 mile or more from a supermarket.” Next, with my goal I started to sketch out my different visualization ideas.
My first idea since I wanted to compare total population to these two variables was to draw out a grouped column chart (figure 3) below.
The second sketch was more representational, I wanted to draw a map of Wisconsin and use dots that varied in sizes as well as labels for the county names and population for people with vehicle access visualized in different counties (figure 4).
My final sketch was a dot chart that compared the total population to the population of homes without cars (figure 5).
After sketching and looking through the data in excel and Tableau Prep Builder, I started to realize that it would be difficult to visualize both ½ mile data and 1 mile data. So, my decision needed to either aggregate the data or eliminate one of the data variables. I chose the ladder because I felt that this would confuse the viewers and I wanted to make sure to keep the visualization simple. My new goal was to visualize: “Number of people without vehicle access that are 1 mile or more from a supermarket.”
I started off testing out each of my ideas with Tableau Desktop. I immediately worked with visualizing bar charts keeping my goal in mind.
After creating the bar charts with the data in Tableau, immediately I found a few issues when trying to compare the two values:
I tried to do some calculations instead to develop a per capita result, however I still had the issue that a viewer would have to scroll to see all the counties. Some things that I felt went well with the visualization was the simplicity and the use of a monochrome blue that helped visual strain.
Next, I tried working with my data by creating a map of Wisconsin. The first visualization I made, I created labels for each county with Wisconsin and values for “Housing units without vehicle count beyond 1 mile from supermarket.”
This visualization used proximity of the counties and placed the dots on their geolocations. Additionally, similarity and closure were exemplified by the whole map of Wisconsin providing a visual cue that the state is Wisconsin. However, there were a few issues, there was no clear visualization of territories as well as there was a lot of whitespaces. Additionally, visualization’s that use shape and scale are hard for users to visualize differences. I continued to play around with the visualization and found out that I could create county borders. I eventually created the visualization below (Figure 9).
The new map had even better proximity and closure as the outlines of each of the individual counties helped fill out the state of Wisconsin. I also decided in addition to using labels to create a heat-map of the data. The colors corresponded to “Housing units without vehicle count beyond 1 mile from supermarket.” I also included the total population for each county for a better comparison of the data.
User feedback brought on a lot of changes:
After feedback I made the changes necessary and created a new column and variable in the dataset called “PerCapita.” I also changed the color gradient to a red. Lastly, I added in an equation for those who are interested in understanding the math from the dataset.
I felt satisfied with the new dataset and decided to go through another round of feedback with different users. The second round of user feedback involved a few changes including my equation (Figure 11) and data labels. The equation did not exactly make sense as I was comparing a total population (people) to number of houses (houses) that do not have a vehicle and are 1 mile or more away from a supermarket. The new equation was:
Lastly, users still felt that the Key was redundant in stating the same information for the heatmap twice. I personally noted in the key that my values did not end at zero or go to a whole round number for the heatmap values. In order to reduce pre-attentive processing, I rounded down the minimum value and rounded up the maximum value in the key. Lastly, I thought about someone that may be outside of the states or not familiar with Wisconsin, so in the title I added in “in Wisconsin.” Pictured below is my final data visualization.
After further feedback It was noted that the per-capita values were too abstract and required further reading. I decided to direct this visualization into percentages to better communicate seen below (Figure 14).
This was my first project where I cleaned a dataset and developed a visualization. I learned a lot about the differences between software tools like Tableau and R especially how powerful Tableau is when creating visualizations. Overall What I learned from this project is to iterate. Instead of designing with your own ideals, consider including stakeholders in decision making process to align your vision as a designer with theirs. Designing in a silo can lead to a lot of unseen issues.
Something that I have been considering is making the visualization more interactive through website functionalities. Essentially javascript micro interactions that would enable a viewer to interact more and gather more data from the visualization. Because I had only so much real estate for a static visualization, I think it would be beneficial to make the visualization dynamic.
