Compare the graphs
Graph A: This graph shows COVID-19 cases in the UK when it first passed 1K at 1,282.
Graph B: How many days more do you think it would take for the cases to reach 10K, how about 100K?
Graph C: Here is another version of the graph, where case numbers are log transformed.
Graph D: Now try make the same prediction again, this time based on the log-scale graph?
How did you go? Which graph (A,B,C, or D) helped the most with your prediction?
In reality, it took the UK
- 11 more days from then to reach 10K confirmed cases;
- and another 23 days to reach 100K confirmed cases
- What do you think? Which graph, the linear-scale graph or the log-scale graph is more likely to lead to a close prediction?
Discussion
Problem
Without proper intervention, case numbers of infectious disease have the potential to grow exponentially. Therefore, log scale graphs seem more appropriate to portray the growth of such cases than linear scale graphs. However, for many, log scale graphs are hard to interpret. For example, in an experiment, participants presented with log (vs linear) scale graphs were less able to grasp the situation accurately and were less worried about a health crisis, partly because log graphs look flatter and more reassuring than linear graphs (Romano, 2020).
Would this help?
This DataViz is an attempt to
Test whether adding reference lines to log graphs can improve interpretation;
Discuss the possible advantage of log graphs at the beginning of a spread:
A. when the rate of exponential growth is relatively stable, and log graphs are therefore not flattened yet.
B. when case numbers are relatively low and seemingly unintimidating, log (vs linear) scale graphs may be more effective in communicating potential consequences, and by doing so, enhance public adherence to prevention measures.
References
Romano, A., Sotis, C., Dominioni, G., & Guidi, S. (2020). The scale of COVID‐19 graphs affects understanding, attitudes, and policy preferences. Health economics, 29(11), 1482-1494.