A2-HeatherDolan
From CS294-10 Visualization Fa08
Contents |
1 - Domain
Public Health/Insurance
I initially was interested in looking at internet and cell phone usage, but the data sets I was finding were either too simple, not that compelling, or required a lot of reformatting. I was also interested in looking at the public health domain as well, so I went with that as I found more diverse data sets.
Initial Question: Are the number of uninsured residents increasing over time in some states more than others? (When posing this question, I had no idea what the answer would be.)
2 - Data
After looking at a lot of data sets and experimenting with a few, both within the selected domain and others, I finally settled on some US Census Bureau data on Health Insurance
http://www.census.gov/hhes/www/hlthins/historic/index.html
It includes 9 years of data regarding how people in each state are insured (government vs. private) or not insured.
3 - Visualizing in Tableau
The data set required several adjustments before it was easily usable. It was available as .xls file and the modifications required mainly involved removing additional text that didn't have data associated with it or groups that were column headings. Date fields also contained some non numerical values that needed to be removed. Once I completed the formatting, it was easy to get started in Tableau.
Initial Visualization
My initial visualization was intended to make sure I had my data correctly sorted out and was able to produce output in Tableau.
It's a simple bar chart of the number of uninsured (that completed the survey) in each state, for each year.
This visualization was not very useful as just raw numbers are used for the visualization and, so states with larger populations and, therefore a larger number of participants, appear to have a larger population of uninsured residents relative to insured residents. Note, I assumed the total number of participants per state were proportional to the states total number of residents.
Iteration 1
For the second visualization, I calculated the percentage of uninsured residents, so that the states with larger populations would not look like they had a larger number of uninsured residents.
It's huge and fairly difficult to digest quickly.
Further Iterations
I then experimented with Tableau and tried several different style of visualizations.
This visualization isn't useful for extracting specific data, but does reveal that, generally, there were not huge changes in the types of insurance, or lack of, over the nine years.
Detailed Breakdown
I tried a few different versions of a detailed display of all of the insurance information for all states for all years.
This allows for examination of specific changes in the type of insurance over the last nine years by state. Note, I've only included two states here, Alabama and Alaska.
In the first, private vs. governmental insurance is encoded by hue, blue and green respectively, and shade encodes insurance type (redundantly as they are also labeled).
However, the question asks about the uninsured and there are labels for the types of insurance, so I removed the color encoding and reformatted in the second attempt.
Map
I also used Map based visualization, where the map's representation of the state encodes the state, the size of the circle encodes the percentage of uninsured residents in the state, and color encodes the year. They years are also overlaid. I excluded some of the years because, with this particular data set, there is so much similarity.
If there was an increase in the number of uninsured, the green circle will have a ring of blue around it if the increase was largest in 2003 and a ring of red if the increase was largest in 2007. Unfortunately, the exact position of the circle on the map doesn't mean anything. I think this visualization is useful for quickly assessing where, if anywhere, there were large increases in uninsured populations.
Conclusion
There have not been large increases in the uninsured population across the country over the past nine years. Values fluctuate from year to year and only a few states have increased from 1999 to 2007. It remains to be seen what will happen to those values for 2008.
For this particular data and now knowing the answer to the question, the map is best method for communicating this information. It's a manageable size and requires a lot less text. A large red area would quickly indicate that a state had a significant increase in its uninsured population. An interactive component that would allow you to drill down into insurance type specific data, as in the case of the detailed breakdown, and compare one state to another would improve the visualization. I wasn't able to explore interactivity prior to posting this assignment, but would like to explore adding an interactive component.
