SG Covid Case Clusters
Presentation for Data Viz SG / HHSG meetup
Topics to be covered
- Introductions
- Why create SG Covid Case Clusters?
- Insights that can be gleaned
- Technology stack
- Steps taken and tried
- Issues, problems, lessons learnt
Hi!
I'm Chi-Loong. Nice to meet you!
I run the Data Viz SG and HHSG meetup groups.
I work in the visualization space. I run my own visualization studio at V/R (www.vslashr.com) doing work for clients.
I also teach data visualization as an adjunct lecturer at SUTD / SIT.
SG Covid Case Clusters
Here's the SG Covid Case Clusters viz at this link (covid.viz.sg)
Why this visualization?
Because we were missing a good dashboard for the public that gives us an intuitive insight into how Covid case clusters are formed.
What's out there
Geospatial marker visualizationscan be useful and can tell you where to avoid going during this period, but it does not give you a sense of how the cases spread and relate to each other.
The Ministry Of Health (MOH) Covid dashboard is good for high level summary statistics, but doesn't go beyond that.
Our media does both summaries and sometimes daily reports but there is often not a visual representation of the clusters.
On visualization
Visualizations fall within a spectrum: Exploratory analysis on one end, and presentation (aka storytelling) on the other.
This visualization is like a visual index of the MOH's covid case notes, and hence would be more of a tool on the exploratory analysis side.
Covid clusters dashboard
Hence, a dashboard, that:
- is far easier to update than manually drawing out the clusters
- allows us, at a glance, to intuitively get case relationship insights
- allows us to search and find each case by ID in the dataset (tedious and difficult on the MOH site)
- allows us to search and find cases via occupation or organization
- allows us to get summary statistics on any subset more easily
Fun questions
- What can we say about the overall distribution of the covid clusters?
- Are vaccinated folks less likely to spread Covid?
- Are vaccinated folks more likely to be asymptomatic?
Technology stack
- Holy trinity of HTML/CSS/JS
- D3.js for all the charts - force-directed graph, and the bar charts
- Bootstrap as the CSS framework for layout and responsitivity
- Lodash.js for data prep and manipulation.
- No JS framework (Vue, React, et al). I almost regretted this (more later)
16th May 2021
23rd May 2021
30th May 2021
6th June 2021
13th June 2021
Issues and problems
- I wish the MOH data was given to us and machine readable. It is not.
- The dashboard doesn't magically occur all at once. It was built piecemeal over time (when I had time).
- Most important to me was making sure the data that comes in every day is transcribed.
- A lot of the functions I buit were so that I could double check on whether I transcribed the cases correctly, e.g. the search function.
Lessons learnt
- If I were to redo this piece, I might have added in a JS framework. It won't make the data entry less tedious, and I don't need the framework for reactivity (D3 has its own data binding), but componentization might have been nice.
- I was planning to do this until HA end, but we got another whole ton of cases again today (June 14th). We'll see.
- Dashboard has some visibility. About 12K+ users since 3 weeks ago, most of it in the 1st week. Not as crazy as elections.viz.sg (this was insane at 300K+ users in 3 days).
Comments + Future
I've open sourced all the code and data, so anyone can play with it. If you use the code and/or data, it would be nice to have attribution! :D
It literally is a dashboard to help dig into the data, so does not clash with what media / researchers do. Please feel free to use!
Any comments, rants, raves on the data viz please feel free to ping me!
We are always looking for speakers for our data visualization meetup groups. Please ping if interested!
Post 2021 (after project closed)
Visualization exploded in the media, because of many people posting on twitter (now X)
Work done for Guangdong CDC