Thread Closed 
Week 3: Where Poor and Uninsured Americans Live
09-11-2014, 02:39 PM (This post was last modified: 09-11-2014 02:40 PM by nbertagnolli.)
Post: #1
Week 3: Where Poor and Uninsured Americans Live
This visualization is a choropleth map representing the poor uninsured adults of the United States who could be affected by changing Medicaid laws. This map is broken down by districts in each state and represents geographic information directly on a map. The number of poor adults who are eligible under a Medicaid expansion as a percentage of the adult population is encoded by a color hue ranging from green to red in increments of 3%. The graphic is slightly interactive, in that it allows the user to highlight all states, states that are expanding Medicaid, and states that are not expanding Medicaid. The visualization is trying to demonstrate that the states who need Medicaid expansion the most are those that are not expanding Medicaid.

I believe that the choice to use hue to encode a magnitude channel was a mistake. Using hue leads to a discrete representation of values. Looking at the graph I can only know the range of values that a particular district experiences. For example, if a district is dark green it is somewhere between 0 and 3%. I think using a color saturation scheme might be more effective for conveying this type of magnitude information. I find it a bit irritating that the dark green represents a 0-3% range but the dark red represents a 12-100% range. Granted it is unlikely that any district will experience extremely large values, but I did find some districts with eligibility in the 20% range. Looking at the hue bar in general I did not expect to see a value that high because each band of color was separated by 3%. If the author was seriously opposed to using hue then adding three more colors to expand the total palate out to 20% would have been nice. This way the magnitude of the data on the higher end could be conveyed effectively. Overall though, since the purpose of the graphic was to convey the idea that the regions with high poverty were not expanding Medicaid this information is still conveyed effectively through the hue bar.

I believe that the interactivity of the graph is unnecessary. It is easy to use and interpret, but I feel that this same information might be encoded in another way. Perhaps the author could have placed a colored boarder around those states that are expanding Medicaid? This would cause those states to pop out, and the others we would know are not expanding the program. This would keep all of the information of the original graphic but make it static so that it could be equally effective in print.

In my opinion this is in fact an effective graphic. I could easily see the regions of the map where there was higher poverty and lower poverty and isolating regions associated with expansions was not difficult and easily visualized.

[Image: uninsured-americans-map.html?_r=0]

The image isn't displaying for some reason. Here is a link to the graphic. Graphic
Find all posts by this user
09-12-2014, 02:53 PM
Post: #2
RE: Week 3: Where Poor and Uninsured Americans Live
I agree that hue should not be used to convey magnitude. Hue should be used for categorical data. Using saturation could have been more informative as we do not need to look at the legend to know the meaning of each color.
Using interactivity to show districts which are expanding and which are not, is a good idea, this way the author does not cram too much information on the same graph (map). Also using saturation to convey magnitude and another color to represent which districts are expanding and which are not might be confusing and would clutter the graph. Further since the author has kept the color coding same for all the three options 'All States', 'Not expanding Medicaid' and 'Expanding Medicaid', it simultaneously conveys regions which need medicaid expansion and their current status.
Keeping them separate strongly brings out the message that the districts/states which need medicaid expansion the most are the ones which are not expanding medicaid.

There seems to be another interactivity option provided which allows to view the same information as color coded spheres sized by numbers. This appears to be a very poor representation to me. Some of these spheres overlap and it is very difficult to estimate the relative size of spheres. The very same information is conveyed by using an overlay and using spheres of different sizes to convey this information was unnecessary.

Overall, I think the using the map style as 'Colored by Share' was effective in conveying the message. The graph could have been better if saturation was used to convey relative ordering instead of hue.
Find all posts by this user
09-13-2014, 12:05 PM
Post: #3
RE: Week 3: Where Poor and Uninsured Americans Live
I think nbertagnolli gave an excellent analysis for this visualization and I agree most of nbertagnolli's points.

However, I do not agree with nbertagnolli on the usage of color hue to encode the uninsured percentage. Actually, I think using color hue is a great choice.
  • Continuous data is accurate and detailed, however, it can also be overwhelming. Sometimes, we need to skip the fine details to capture the "big picture".
  • The author used interactivity to facilitate a "hierarchical scheme" to give the value of the uninsured percentage. First, we get a coarse idea of the values in terms of ranges from the colors. Then if interested, we can use interactivity to pop out the details.
  • Imagine instead we use a continuous color map. We can visually tell the difference between two colors if the underlying value are quite different. However, this would be quite difficult if the two values are close, for example, the values in North Dakota. In the later case, our mind will implicitly group these values into a group and think they are "close", which is equivalent in effect to using a discrete color map.

As pointed by nbertagnolli, the description of the color map is inaccurate. Most of the colors span 3% while the rightmost one (red) represents a much larger range: >12%. However, this is not indicated in the visualization. This could be misleading. For example, if we glance at Alaska without the details popped out, we would think the uninsured rate for northern and western Alaska to be between 12% and 15%, while in fact it is 19.6%.

I agree with MukteshKhole that the usage of interactivity to switch between Medicaid Expanding/Medicaid Not Expanding is very good. The three maps are exactly the same except that different regions are highlighted. This is very easy for us to interpret. On the other hand, given the various size of the states, the local data-ink ratio would be very high in some parts of the visualization and very low in others, which will make the interpretation hard.

Finally, as pointed out by MukteshKhole, the MAP STYLE "Sized by number" is very misleading. It looks like the author is trying to present the same information in a different way. If that is the case, then the "Colored by share" choice did a much better job as explained by MukteshKhole. On the other hand, do the locations of the circles represent anything? For example, the location the sample was taken? This is no clear from the visualization.
Find all posts by this user
09-14-2014, 10:22 AM
Post: #4
RE: Week 3: Where Poor and Uninsured Americans Live
I agree with lzhang's opinions for the most part. The user could select three tags above the map to get a general ideas of all states, not expanding medicaid and expanding medicaid in different states. Meanwhile, user may select a place to check exact number of poor and insured residents and share of eligible adults of that place. These interactive ways are good. I also like the color hue in this visualization since it is very intuitive. The green and red color are good options because it is easy to tell regions of share of eligible are less than 3% or more than 12%. I have a strong visual impact on these two colors.
It would be great if the designer would present more clear on the legend of "Map style" since I was confused by "colored by share" and "sized by number". It wasn't clear in this visualization.
Find all posts by this user
09-14-2014, 11:05 AM
Post: #5
RE: Week 3: Where Poor and Uninsured Americans Live
Personally , I found this a great visualization. When we are representing a lot of numerical data such kind of visualization has a great aesthetic appeal and does not clutter the map with lot of numerical data. At the same time all the required information can easily by obtained by clicking on individual points.

Few key points, which I thought could be improved in the visualization, are as follows

1. Colors: As the visualization shows a comparative study about the states who do and do not participate in the Medicaid. I thought choosing 2 different color schemes to represent the data showing the share of eligible adults would have greatly helped. By having two separate legends for states who do and do not participate would greatly help people who want to identify the regional data at one glance .

2. Inconsistencies: While switching between the three tabs, I noticed that the color scheme to represent the numbers seems to be inaccurate. For example in the non-expanding Medicaid tab. It shows role of people who are ineligible for insurance as 15 percent with light orange color that according to legend should only indicate 6 to 9 percent.
Find all posts by this user
09-14-2014, 11:48 AM
Post: #6
RE: Week 3: Where Poor and Uninsured Americans Live
I feel the graph effectively conveyed what it meant to, i.e. the states are not participating in an expansion of Medicaid are home to a disproportionate share of the nation's poorest uninsured residents.

What I liked about the graphic:

1. The separation of graph into 'All states', 'Not Expanding Medicaid' and 'Expanding Medicaid' makes it easier to interpret the data.
2. The graph is not cluttered, given the amount of data that it represent.

What I did not like:

1. The second graphic provided on selecting 'Sized by number' was confusing and repetitive. I did not understand the need of this representation. Moreover the data that was represented in this way was cluttered and did not convey what it should have.

Overall, the graph was a good visualization and the data provided could easily be interpreted by the user.
Find all posts by this user
09-14-2014, 12:49 PM
Post: #7
RE: Week 3: Where Poor and Uninsured Americans Live
I actually really like the use of interactivity in this visualization. While I agree with the comment that it would be possible to encode the information without the need for separate tabs (e.g. through shading borders based on which states are expanding/not expanding Medicaid, as was originally suggested), I think that specifically highlighting certain regions while the others fade out makes it easy for viewers to focus on data that is relevant to the information this graphic is trying to present. If it was all presented at once, I think it would potentially be quite difficult for a viewer to avoid conflating neighboring regions.

I also think this visualization is a (mostly) good example of using color as an encoding channel without it being overwhelming, namely by limiting the possibilities to five discrete categories. This has the downside of losing some of the data's granularity, but the interactive element helps make up for this by providing the exact values for each region by hovering over them. I definitely agree that saturation would be a better channel than hue for encoding these continuous values, though. The current scale looks like it uses two color schemes diverging from the 6-9% category, but using 0% as a baseline would make more sense in my opinion.

I also am intrigued by the point raised about the extreme magnitude of the final color category: a range covering all values greater than 12%. This once again causes this visualization to lose a great deal of granularity, but I think the creators of this visualization didn't want to emphasize the differences above this point: rather, they used it to allow viewers to easily classify all of the regions above this threshold into a mental category of "most in need of Medicaid expansion," which helps viewers easily see how the graphic relates to the associated story.
Find all posts by this user
09-14-2014, 03:26 PM (This post was last modified: 09-14-2014 03:27 PM by zhiminl.)
Post: #8
RE: Week 3: Where Poor and Uninsured Americans Live
I think nbertagnoli do a very good analysis in this visualization, especial the using of color in this picture. The people before cover the most efficient part of this visualization so I just add some comment base on their discussion.

First of all, We already have 5 color in this density visualization, so may be add more color in this visualization won’t be a good plan.I am total agree that the red represent from 12% ~ 100%(technically) is a bad idea but I solution to this problem is resize the percentage of each color. Instead of assigning each one with 3%, we can do it with 5% or a little more.

I Agree with Mattew’s opinion that interactive technique is used pretty well in this visualizationBig Grin. Add additional colored boarder to this picture may be a good plan but first like I say before, this visualization has high density so I think may be we shouldn’t add more information into it. Secondly, I pretty enjoy using the zoom in and zoom out technique which give me more detail about a tiny region in the map.

I like lzhang’s idea that “capture the big picture”Idea, and I believe that’s the main purpose that the author create this visualization.
Find all posts by this user
09-14-2014, 03:57 PM
Post: #9
RE: Week 3: Where Poor and Uninsured Americans Live
I think this visualization is actually a effective one in its color map choosing and pop up effect.

About its color map, I don't think they simply use hue to represent uninsured percentage. I would rather take it as a diverging color map, with 6%-9% as the base line, below 6% is greener as decreasing and above 9% is redder as increasing. Because the average percentage in all states are 8% which is between 6% and 9%. And the pop up effect when clicking 3 tabs is useful for people who are concern about the data in those specific areas.

However, one thing about their colored by share graph is that people may be misled by the size of different regions and take the size as the number of poor and uninsured. I think that is why they create another graph that use size of spheres to show the exact number. But I do not think combine two channels(i.e. size and color) to represent different attributes(i.e. the number and percentage) in one graph is a good idea. Because people may not focus on two different channels at the same time and then get confused.
Find all posts by this user
09-15-2014, 04:35 PM
Post: #10
RE: Week 3: Where Poor and Uninsured Americans Live
This visualization uses a diverging color scheme, but the there is not logical pivot point that I can see. I can see no good reason to make the ranges between 3% and 9% light green and light gray. This is one of the problems we talked about with color palettes. The intensities should have reinforced the data (light intensity = low percent, darker intensity = high percent).

I have trouble analyzing this graphic objectively beyond this simple point because it is very politically charged. The use of harsh phrases like "Republican-dominated" and "stranded without insurance" clearly indicate that this visualization was intended to embarrass Republicans into expanding medicaid. This leads me to believe that the designer may have chosen to violate the expressiveness principle with color choices that might mislead.
Find all posts by this user
Thread Closed 

Forum Jump:

User(s) browsing this thread: 1 Guest(s)