Thread Closed 
Week 11: RFID Tags Track Possible Outbreaks
11-06-2014, 05:54 PM
Post: #1
Week 11: RFID Tags Track Possible Outbreaks
Here is a interesting visualization from scientific american. European researchers tagged people in a hospital with RFID tags and tracked their movements. Who the people interacted with is displayed in the visualization which can be found below:

RFID Tags Track Possible Outbreaks

The visualization splits people at the hospital into groups such as nurses and patients. Each person is represented by a bubble where each bubble consists of an inner and outer bubble. The inner bubble's area represents the number of people of a particular group that person has interacted with and the outer bubble's area shows the total number of people that person has interacted with. Individual interactions between two people are represented by a line linking two bubbles together. The visualization is interactive and different groups can be selected.

Things I like:
1. Color, grouping, and relative size are all used to show group membership. I think that the use of three separate channels for this data really made it apparent which people were in which groups.
2. There is a easy to find button that describes how to interpret the visualization.
3. A lot of data displayed in a really small area.
4. Interactive features make it easy to study interactions between different groups which should make it easy to implement policies for those groups. (example nurses should only interact with a small group of patients)

Things I didn't like:
1. The bubbles overlap in the visualization which makes it hard to tell the actual relative size of each individual bubble. For the nurses it is so bad that the entire group of people look like one long rounded curvy triangle.
2. There is no key for the bubble size or indication of the scale that the bubble size is on. Other than counting the number of lines coming out of each bubble there is no way to judge the connectivity of a person.
3. I think the hue choice for nurses and physicians is a bit too close to be easily differentiable.
4. The visualization seems a bit slow.
5. It seemed like the lines could be spread out more evenly throughout the visualization instead of concentrating them into the center. This would cut down a bit on the occlusion.
6. Lines change color halfway through which makes following them kind of hard.
Overall a pretty cool visualization.
Find all posts by this user
11-09-2014, 03:44 PM
Post: #2
RE: Week 11: RFID Tags Track Possible Outbreaks
Overall a very good critique. This visualization does a very good job of showing the numerous connections within the dataset. I agree that they could have spread the bands out a little more so we could see more individual lines and prevent some of the overlapping occlusion especially considering there is a lot of untouched white space to utilize.

I agree the overlapping circles detract from the visualization since they overlap and don't have a clear cut legend explaining their size.

Overall though it is an effective visualization because it shows the most important hypothesis of the dataset, that nurses should be the most concerned about hygine because they have the widest variety of interactions in the hospital environment.
Find all posts by this user
11-09-2014, 09:03 PM
Post: #3
RE: Week 11: RFID Tags Track Possible Outbreaks
I agree with all of the points in the critique. I think that changing the color for the nurses to a brighter green or something would help a lot. It would be nice if the highlighted group expanded so that the bubbles didn't overlap so much. I think the visualization would be better if it was quite a bit bigger. Then a lot of the issues would be solved.

On the whole, I think it was a very interesting visualization. I was surprised by many of the different interactions between the different groups.
Find all posts by this user
11-10-2014, 01:47 AM
Post: #4
RE: Week 11: RFID Tags Track Possible Outbreaks
I think, this kind of visualization is called chord diagram. They are really effective in showing interaction between different groups in large datasets. Overall in this context, the selection of this visualization, the color scheme for denoting different groups are very efficient.

Like others I do agree that overlapping bubbles of individuals in all the categories is a negative thing. But this is mostly due to the dimension of the circle for the visualization. I believe adding another level of interaction to zoom in and out would have easily alleviated this issue. Furthermore, that could have made it possible to visualize much more information.

But maybe for this context the level of detail presented is enough.
Find all posts by this user
11-17-2014, 04:54 AM
Post: #5
RE: Week 11: RFID Tags Track Possible Outbreaks
Before I get started, just as a side note, it is interesting to note that there are some patients who never saw a doctor, a nurse, or a ward assistant (and two who saw nobody at all). Smile

I agree with the overall consensus - the visualization is quite effective. I also agree that encoding the number of interactions per person by area was a poor choice, or at the very least, not useful (I can't think of a better way, without labels).

I also don't think the banding is much of a problem. Each individual interaction isn't what's important, but rather the overall pattern of interaction between groups. This thought leads me to less of a visualization question, and more of a question about their hypothesis. Is categorizing people based on their group the best way to assess transmission risk, or even patterns of interaction? For example, note that nearly half of the doctors never interact with any patient (on the other hand, maybe these work part time, or something).

Anyway, with regards to suitability for intended purpose, I think the original critique was spot on.
Find all posts by this user
Thread Closed 

Forum Jump:

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