A2-JeffreyPatzer

From CS 294-10 Visualization Sp10

Jump to: navigation, search

~ Data for this visualization was pulled from the following website's page - http://www.guardian.co.uk/news/datablog/2010/jan/14/haiti-quake-aid-pledges-country-donations - which presented the data stored in a google doc - http://spreadsheets.google.com/pub?key=tMozJlQiN8ftSj0s8tLx31g&gid=0 | Data may have changed since the time of creation for this visual. ~


The visualization featured above is trying to answer two questions. Which countries are providing the most money (as a percentage of the whole)? And which countries have the most giving people (based on population)? Due to the constraints of the data, these were the two questions that could be attempted to answered the most honestly. I created the visualization in photoshop because it gave me the most flexibility for getting rid of chart-junk and unimportant factors.


I focused primarily on using the single sheet of data to create the graphic. I also wanted to try and mimic Tufte's visualizations by making the visualization clean and sparse, but also detailed.


I chose to use a combination of a bar and scatter plot to encode the data. The bars are simply thin lines, and the scatter points faded dots. The bars represent the donated per person dollar amount per country. The dots represent the funding committed percentage per country. According to the data set, the country amount can be assumed to be funding being provided by a government, not including its private citizens. While additional categories were present that listed individuals and various emergency funds, these did not allow for a country comparison (dollars per person) and are therefore included in a side table on the visual.


The reason I chose to use the bar and scatter for this data is it allows for quick comparisons. While the exact numbers are important for financials, the quick comparison between countries seemed more important when trying to answer the question of who is "doing more".


The bars use the visual encodings of both length and position. The use of position and length seemed the most appropriate to compare the quantitative data, based on Mackinlay's rankings. I also tried to keep the data-to-ink ratio low, so I made bars into lines and also alternated the shading of the bars slightly to help with not creating white-space bars. The position of the bars is important because it represents relatively how much a country has given. This encoding provides perspective for the per person dollar amount. Additionally, since I wanted to emphasize overall and quick comparison, I have no chart lines or explicit y-axis. They seemed to be chart junk and detracted from data comparison. I believe using these line bars allows one to quickly pick out the countries that have given a large amount relative to their size, whether or not the sum itself is large. It allows one to quickly sort and pick out the leaders and laggers in this category.


The dots seemed appropriate to use to describe the total amount donated. They simply use position along the y-axis to encode their value. This seemed to work the best to emphasize the differences between the total amount of donations by each country. The dots allow for a person to easily see the invisible scatter plot trendline, which reenforces the ordering of the countries along the x-axis.


I chose to stick to a gray-scale color scheme because I didn't want the chart to be confusing and hard to view. The amount of data was not huge by any means and the story I wanted to tell did not require color differentiation. I at one point contemplated using flags as country encodings, but figure that it would add too much color variation and that not everyone knows what flag corresponds to which country. Rather I chose to use country codes to encode the country.


For scale, I created the graphic on a 8.5 X 11 canvas because it was easily to manipulate a larger canvas, objects, bars, and so on. However, I designed the visual with the idea that it should be able to be shrinked/expanded and still allow for the data trends to be picked out. I have the x-axis country codes evenly spaced so that the bars and dots remain the same distance apart. I encoded two variables on the y-axis that are not related and are different enough that they allow for the data to be separated. The y-axis values are scaled to contain the greatest country value needing to be represented, but scaled in such a way that the per person dollar amount it able to fairly accurately represented.


Additional donors were listed as well that did not fall into the country category. Since those numbers provided excellent percentage comparisons, they did not have a per person stat. Additionally, they contained lumpings of groups that could be considered to have been placed within a country. The data was not clear on the breakdowns for those sections. I decided to omit those from inclusion on the bar graph, but still included them within visual as a small table. I figured this data simply help to complete the picture of aid being provided, but did not necessarily need to be included in country data. Since the values had much greater variation than the country data, it would have greatly expanded the size of the visualization and distorted the scale, resulting in less information for things lower on the scale. While comparing these values with the countries does require some work, excluding them from the bar and dot section allows the trends and story to be more effectively display (or at least I think so).


Some concerns and drawbacks to my design are listed below:

  • I think the graph might at first be construed as a weird lollipop diagram, rather than two sets of statistics.
  • I could not think of a good way to encode population. While I entertained the idea of changing the dot size, I could not think of a good way to proportionally do this. Do I make the size of the circle proportional to the entire world's population? I also figured it might make it more difficult to read trends in the graph so I decided to withhold population information from the graph. It might have been possible to encode it along the x-axis by ordering the countries according to that stat, but the trend of the dots would be lost.
  • I was hesitant to include two separate scales on the y-axis since they might cause some confusion or misunderstanding when comparing the data. However, layering the y-axis allowed for more data to be included and for a more important understanding of the data to be revealed. WIthout having both the bar and dot pieces of the visualization, it would not be possible to see that Canada pretty much dominates both categories.
  • Having to include a table for non-country data. While I ultimately think it was the best decision for the data provided, it is still a bummer it could not be more appropriately encoded within the bar and dot sections.
  • Bars being created by the white space rather than the lines. I tried to negate this effect by centering the bar over the country code, which meant the white space did not line up in a way that the viewer might read it that way.
  • By choosing to emphasize comparison of data, rather than actual numbers, the actual numbers do not get encoded within the data.
  • Even though this visualization was made within the past week, the data has already changed and altered, making this visualization out-dated. Due to the method of creation, altering the data would not be easy to do. So while the graphic is better than you might get in Excel, the ease with which one can change it is much more limited.
  • The data provided does not provide GDP earnings and a lot of other data that might help provide a much better context for the data.
  • Dots are gray-scaled. To emphasized them more I might make them red or another vibrant color to help them stand out more.


[add comment]
Personal tools