Read on Twitter
After reading about pros and cons, I am pretty convinced that graphs and stats about #COVID19 should be normalized per capita or reported on comparable regions... a [THREAD] 1/15
First off, thanks to everyone who chimed in yesterday with links and opinions https://twitter.com/xamat/status/1260463107178184704?s=20.">https://twitter.com/xamat/sta... Also, for the purpose of this discussion I will focus on reported deaths, since I think that is much more robust figure than infections, which greatly depends on testing 2/15
Next, let& #39;s start with the practical aspects: why would I care about this? I get daily messages from my friends and family in Spain worrying about "how bad" things are in the US. They are reacting to reports on overall number of deaths or graphs like this one 3/15
They are unaware that California, with a population comparable to Spain, has exactly 1/10th the number of deaths of my home country (2700 compared to 27000). You may think my friends & family are unsophisticated. I think that is a side effect of reporting raw numbers 4/15
I understand normalized numbers can be interpreted as benefiting the US and can be used politically to defend the response from the US government. For the record: I think this administration and its clown in chief has been an absolute disaster handling the pandemic 5/15
However, it is also true that some states, including California, have done a much better job, and that should reflect on the overall stats for the country 6/15
So, what are folks arguing for not reporting per capita COVID figures? The least sophisticated concern is that "a life is a life", and reporting per capita lowers the value of a life in larger countries 7/15
However, the opposite can be argued. Not normalizing assumes that a life saved in California is worth 1/10th of a life saved in Spain. The "value argument" is flawed, and comparing unnormalized figures only makes sense for regions of comparable populations 8/15
Let& #39;s now look a bit into what "experts" are saying on this. Here is the great @jburnmurdoch explaining why they chose to go with raw numbers in the awesome @FT infographics https://twitter.com/janinegibson/status/1244519429825802240?s=20">https://twitter.com/janinegib... (skip to minute 2) 9/15
This is described as a "judgement call". The goal is, according to John, to favor immediacy and "visceral reaction" to raw counts. That, has nothing to do with data or science, and as I showed in the intro, it leads people to *wrong" visceral reactions. So, I don& #39;t agree 10/15
It is great to see that @FT& #39;s interactive graph now does provide the option to normalize by millions. https://ig.ft.com/coronavirus-chart/?areas=usa&areas=esp&cumulative=0&logScale=1&perMillion=1&values=deaths">https://ig.ft.com/coronavir... 11/15
An important observation from the graph above: it does not make the US "look good". While it& #39;s true that the overall impact of COVID in Spain has been larger, it& #39;s also true that Spain& #39;s trajectory is looking much better than the US recently, and that is clear in the graph 12/15
Finally, @CT_Bergstrom has another interesting thread defending non-normalized scales (thanks @flaviovdf for the tip) https://twitter.com/CT_Bergstrom/status/1249930293928030209?s=20">https://twitter.com/CT_Bergst... 13/15
It is important to point out that that whole thread is focused on why normalization does not work when using a linear scale. As Carl points at the end of the thread, log scales avoid this problem. But, I will go further and say that log scales plus normalization are best 14/15
I will also agree with @jburnmurdoch& #39;s comment on that thread that comparisons would be much better done at region (not national) level, but will add that those regions should have comparable population (e.g. California/Spain)
https://twitter.com/jburnmurdoch/status/1251589773040418818">https://twitter.com/jburnmurd... 15/15
https://twitter.com/jburnmurdoch/status/1251589773040418818">https://twitter.com/jburnmurd... 15/15
I hope this thread was at least informative. Looking forward to the comments.
