E-learning Python for Ocean Mapping (ePOM) project.
Complementary slides to the Data Visualization module (part of the Introduction to Ocean Data Science training).
More details at https://www.hydroffice.org/epom
2. WHY DO WE NEED DATA VISUALIZATION?
“Computer scientists are going to have to realize that
primary memory is the human brain, not RAM”
(Buxton, 2001)
AMOUNT OF
AVAILABLE DATA
HUMAN COGNITIVE
ABILITIES
Time
3. WHY DO WE NEED DATA VISUALIZATION?
“We are all cognitive cyborgs in this Internet age
in the sense that we rely heavily on cognitive
tools to amplify our mental abilities.”
(Ware, 2010)
“Often the most effective way to describe, explore,
and summarize a set of numbers – even a very large
set – is to look at pictures of those numbers.”
(Tufte, 2001)
4. CRITERIA FOR DATA VISUALIZATION
Perceptual hierarchy of visual cues
(Cleveland and McGill, 1985)
Accuracy
LENGTH (ALIGNED)
LENGTH
SLOPE ANGLE
AREA COLOR INTENSITY
COLOR HUE
VOLUME
5. CRITERIA FOR DATA VISUALIZATION
Which chart type?
Try with different ones!
Example of chart-chooser → Abela (2009)
6.
7. CRITERIA FOR DATA VISUALIZATION
Which colormap?
Think at the following color wheel …
(source: Wikimedia Commons)
10. CRITERIA FOR DATA VISUALIZATION
𝑫𝒂𝒕𝒂 𝑰𝒏𝒌 𝑹𝒂𝒕𝒊𝒐 =
𝐷𝑎𝑡𝑎 𝐼𝑛𝑘
𝑇𝑜𝑡𝑎𝑙 𝐼𝑛𝑘 𝑖𝑛 𝑡ℎ𝑒 𝐺𝑟𝑎𝑝ℎ𝑖𝑐
(Tufte, 1983)
vs
???
(data source: http://pypl.github.io/PYPL.html)
11. CRITERIA FOR DATA VISUALIZATION
𝑫𝒂𝒕𝒂 𝑰𝒏𝒌 𝑹𝒂𝒕𝒊𝒐 =
𝐷𝑎𝑡𝑎 𝐼𝑛𝑘
𝑇𝑜𝑡𝑎𝑙 𝐼𝑛𝑘 𝑖𝑛 𝑡ℎ𝑒 𝐺𝑟𝑎𝑝ℎ𝑖𝑐
(Tufte, 1983)
Experiment on Data Ink Ratio
(Inbar et al., 2007)
• Approach: 87 students rated 2 graphs from Tufte (1983) work.
• Findings: a clear preference of non-minimalist bar-graphs.
• Take away message: “People did not like Tufte’s minimalist design of bar-
graphs; they seem to prefer "chartjunk" instead”.
14. • Well-tested, popular tool → First release: 2003
• Designed like Matlab → Ease the switch from Matlab
• Many rendering backends → Cross-platform, multiple formats
• A major weakness is the rendering speed for large data → Slow!
• Able to create just about any chart (with some efforts)