Thread by @ProfDFrancis

Here's something impressive.

A mathematical model that tells us that Germany's wave of Corona will finish at the end of May.

Here comes the science bit.

Lots of Greek symbols.

Must be good.

And further complex and impressive stuffs:

(Hat tip to Michael @Mielewczik for sending me this and pointing out this key bit.)

So this is great.

Or is it?

How does the math work?
In fact it is less painful than it looks.

Carl Friedrich Gauss was perhaps the world's greatest mathematican. His name is given to the Normal distribution, amongst many other things.

Epidemics look like Normal distributions, but surely that can't be it?

Oh dear.

So now we can see what happened.

They noticed that epidemics look like a Normal distribution.
And then said, "OK then, let's model it with a normal distribution"

Unfortunately in a real epidemic

When I wondered this a few weeks back, why do epidemics look vaguely Normal, I independently had the brilliant flash of insight, unrelated to receiving a whatsapp from an Orbita-HQ PhD fellow "google for: wiki SIR model"

... that epidemics DO look vaguely Normal, if you are pretty lax about what is Normal.

Their onset is

... because they have to be that shape.

When there is hardly anybody with it, there is little opportunity to give it to new people

Their offset is

Because both start and end are pretty much obliged to be sigmoid, because of the nature of biology, and the math of transmission, we are on the starting point of being Normal.

The only question is how the the upstroke and the downstroke know that they must be symmetrical with each other?

This is the problem.

It only looks Normal, and only vaguely, in the sense that any biological uppy-downy curve looks Normal.

The thing we care about, i.e. the symmetry, so that we can predict the latter half from the first half, doesn't exist.

The good news from all of this is that it means we don't have to get into the Greek letters and things.

We can do a broad estimate by eye, with only the same degree of wrongitude.

Have a good day and stay safe, my friends!

DEFINITELY 110% agree

What I was trying to show is that we can vaguely extend the graphs that are almost back down to zero, and probably be reasonably successful.

But in the ones where we are hopefully somewhere around the top (but pre or post, who knows?) we DO NOT KNOW.

That's why I recommend SHARPIE SCIENCE as a sense check.

When you are nearly at the bottom of a staircase, you can make a reasonable guess how the next few seconds of your life will go.

But when you are far from the bottom, the future is much more unknown and unknowable.

Newly updated meta-analysis!

(If PHE should correct its small errors, I should correct my big ones)

I have no idea where to draw the first 2 curves ... err ... I mean I have used a 7th Generation Neural Network with Hypercooling Superthreaded Monte Carlo simulations: