# 2020/05/21 Science for COVID-19

Let's talk about science.

One of the most frequent questions I get as a scientist is "What does a scientist do?". In short, we extend our basic logic to something we don't know yet. Let me give an example. How much apple is produced in Germany in a year? I presume you don't have an immediate answer. But you know how much apple you eat yourself, like in my case I eat maybe 1 apple a month and maybe a bit of apple juice. That makes around 500g of apple a month, and probably around half of it comes from other countries, like Spain for example. That makes 250g a month, i.e. 3kg a year. Multiplied with the number of inhabitants (86,000,000 in Germany), it gives 208,000 tonnes. And I know that around half of all food is wasted, and probably during the production process a lot more gets wasted. So we can multiply it by 3 and we get 774,000 tonnes, which is essentially my final answer. According to Wikipedia, Germany produced 1,032,913 tonnes in 2016 and 596,666 tonnes in 2017. So my estimation of 774,000 tonnes is quite good.

While this kind of guess is quite rudimentary, it often gives a useful insight that we wouldn't be able to come up with intuitively. What's important here is that whatever science tells, what counts is the reality. In the apple example I gave above, science says 774,000 tonnes, but the reality is 596,666 tonnes and in this case we can safely discard what science says. In this sense, what I'm going to talk about below is a set of my estimates, which are certainly better than knowing nothing, but as soon as the real numbers are available, the numbers below should be discarded. But one way or other, I'm also going to describe my considerations, so I hope you'll know how reliable and where it could potentially be unreliable.

## SEIR model

The first idea I got was based on what's called Law of mass action, which I thought could describe the number of patients over time. This one consisted of three variables: 1, *Susceptible* (uninfected ones in the society); 2, *Infected* (and potentially infect others); 3, *Removed* (either healed or dead). I then did a calculation and posted the results on Facebook. While posts on Facebook often don't bring anything, this time I got a piece of useful information: my model is basically called SIR model but epidemiologists use this model called SEIR, which includes one more variable, *Exposed*, which is for the infected ones in the incubation time, i.e. infected, but no symptoms and don't infect others yet. I don't know exactly whether this one variable plays an important role (at least as far as I could see, it doesn't), but this is a particular variable that probably characterizes the current coronavirus, since we have a particularly long incubation period.

I found out that there was this website called data.europa.eu where you can download the number of infections for countries around world. Based on this, I could compare the results between the reality and what it could have been like without lockdown in the last 3 months.

$$\frac{\partial S}{\partial t} = -\frac{R_t}{T_{\mathrm {inf}}IS$$