Using a logarithmic scale to represent COVID-19 growth

[scottdave]

TL;DR Summary

I came across this interesting thread showing the difference between linear and logarithmic scales when visualizing coronavirus infections.

The author, John Burn-Murdoch, shows here ( https://threadreaderapp.com/thread/1237748598051409921.html ) how the logarithmic scale can give a better "sense" of what is happening. In linear scales, some countries' data is squashed to almost nonexistent, while others explode out of control.

I wasn't sure where the best place to post this - I thought it was interesting and wanted to share.

 

[phyzguy]

I agree that the log scale is more informative. This site has a lot of good information, and allows you to toggle back and forth between linear and log scales. Note that on a log scale the total cases outside of China has been growing exponentially since February, and shows no sign of departing from a straight line (exponential growth).

 

[Chestermiller]

I agree that the log scale is more informative. This site has a lot of good information, and allows you to toggle back and forth between linear and log scales. Note that on a log scale the total cases outside of China has been growing exponentially since February, and shows no sign of departing from a straight line (exponential growth).
View attachment 258928

I'd be interested in seeing what this looks like on a log-log scale plot. Any chance you would consider doing that?

 

[phyzguy]

I'd be interested in seeing what this looks like on a log-log scale plot. Any chance you would consider doing that?

I don't actually have easy access to the data to re-plot it. Why are you interested in that? Exponential growth should be a straight line on a log-linear plot, and it looks like that is what we are seeing.

 

[Chestermiller]

I don't actually have easy access to the data to re-plot it. Why are you interested in that? Exponential growth should be a straight line on a log-linear plot, and it looks like that is what we are seeing.

I don't know why. I just like to look at the data in different ways, and I though that it would even be straighter on a log-log plot.

 

[mfb]

A logarithmic date axis? It doesn't even have a well-defined zero to use.

It won't make anything look straight. It will compress the time where the cases explode even more.

 

[Chestermiller]

A logarithmic date axis? It doesn't even have a well-defined zero to use.

It won't make anything look straight. It will compress the time where the cases explode even more.

I’d still like to see what the log-log plot looks like.

 

[FactChecker]

I’d still like to see what the log-log plot looks like.

By putting both axes on a log scale, the general shape should go back (from the log-plot) to the exponential shape. Both axes of the original plot will be distorted similarly. But I think that the log of the time axis will be hard to use or interpret.

 

[willem2]

I’d still like to see what the log-log plot looks like.

I got the numbers from the source code of the webpage.

All logs to base 10. Note that the point where the growth starts to increase at 1.47 or so is at about 20 february

 

[Chestermiller]

I got the numbers from the source code of the webpage.
View attachment 258972
All logs to base 10. Note that the point where the growth starts to increase at 1.47 or so is at about 20 february

Thanks very much. I was most interested in the part between 1000 and 300000 cases.