Hi,
"All models are wrong" is an old saying in statistics, attributed to statistician George Box.
It holds true in the corona crisis. Epidemiological models are by definition a simplification of reality and therefore can’t be regarded as absolute truth.
That’s, however, not always properly understood. Take this headline from the Financial Times: "Coronavirus may have infected half of UK population — Oxford study".
A new epidemiological model supposedly shows that the vast majority of people with the infection hardly get symptoms. Half of the UK population may already have had the coronavirus! Grist to the mill for people who think all these measures are exaggerated.
But, if you look up the Oxford study, you see: it’s a theoretical exercise. The scientists investigated how sensitive an epidemiological model is to different assumptions.
If "rho" – the proportion of people who end up in hospital – is small enough, it could mean that up to 68% of the British population have had coronavirus. Could mean. It could also be a small proportion, depending on the assumptions.
As the first author said on Twitter: "These are not forecasts."
But some models are useful
But the aforementioned saying has an addition: "All models are wrong, but some are useful."
If you keep the limitations in mind, models are important for understanding what’s going on. So in the past week, I’ve been reading about epidemiological models.
At its simplest, we’re dealing with a so-called SIR model. Part of the population is Susceptible, part is Infected and part is Recovered. Although that R can also mean that someone passed away as a result of the disease, which is why it’s sometimes called Removed.
Take those three groups and add some assumptions, for example how contagious the disease is, how long the incubation period lasts, etc. This allows you to perform a fairly simple simulation. On the website Melting Asphalt you can play with such a simulation, by using sliders to adjust various assumptions.
There are more variations of the SIR model to consider. It’s possible that you are a carrier, but do not yet show any symptoms. If you add Carriers, you have the SCIR model.
Or you can assume that someone can become reinfected after they had the disease, comparable to the STI gonorrhoea. Then you no longer have an SIR model, but an SIS model.
To get a better feel for the possibilities of such models, check the video from 3Blue1Brown. As the maker himself emphasises, it’s a simplified model. But it gives you a good glimpse into the different assumptions that are being made.
#coronafree
I usually share a #NerdAlert here. But maybe it’s better to make it a coronafree corner in the coming weeks. I don’t know about you, but I sometimes crave something that doesn’t have anything to do with the current crisis.
So I’m looking for nice tips, unrelated to coronavirus. Like Maurits Martijn’s profile of Evgeny Morozov, which was published on The Correspondent last week. If you ignore the first paragraph (because it mentions corona), it’s a great #coronafree read!
Morozov is one of the most important technology critics in the world. Last year, he launched The Syllabus, an online platform that breaks with the laws of the attention economy. On it, he and his colleagues share the best articles, podcasts and videos. Stuff that often doesn’t pop up on your timeline because it hasn’t necessarily gone viral.
Maurits joined Morozov last spring for the platform’s launch and wrote a very enjoyable piece about it.
Do you have #coronafree tips? Share them in the contributions.
Before you go...
... the webcomic xkcd by cartoonist Randall Munroe is fantastic, also in times of corona (thanks to colleague Tim Strijdhorst).
Prefer to receive this newsletter in your inbox?
Follow my weekly newsletter to receive notes, thoughts, or questions on the topic of Numeracy and AI.