Articles to 2020-07-18

Zum Seitenende      Übersicht Artikel      Home & Impressum

First the link to this week's complete list as HTML and as PDF.

***

Nothing Adam tells us is new or surprising. R is an important biological parameter for modelling epidemics but meaningless for a purely descriptive statistic of what is going on. Besides the slope of rising or falling trends – a slope is not the same as short wiggles – it also needs the generation time of transmission. Different people use different generation times without telling us or, like the RKI, change them in middle of a data run without saying so.

The main point here is that graphic representations of the data tell us absolutely nothing unless drawn with a logarithmic scale. Given that, R really is the only meaningful statistic politicians are ever given. It's not their fault sticking to the only thing they have.

So called “nowcasting” is an utter fallacy. With data being delayed for up to two weeks, all you can tell about today is only a forecast, nothing more. It won't and can't reflect changes of the situation before the data are in[1]. As such it is a pernicious form of disinformation.

***

Understanding mathematics as in Lammers et al. is one thing. Nobody can reach meaningful conclusions without being told the data. In linear diagrams you can't possibly see any exponential trend before it's far too late, whatever your education. Have you ever seen diagrams for Covid scaled logarithmetically anywhere? If you consistently lie to people and withhold all meaningful facts from them, don't be surprised if they misjudge the situation. It's not their fault.

***

Alright, so psychology has only taken a measly four decades to correct its record as O’Grady mollifies us, and now this Cargo cult can continue playing at being a science, right? I wonder how much of that is driven not by Eysenck being wrong and having faked his data but by his results being seen as politically incorrect by today's Ministry of Truth. After all most of what Freud has written is known to be either opinion or fabrication but it is what his adherents want to hear, so who cares?

***

Siegenfeld et al. talk a lot of sense about what models are, what they can and can't do, and what makes a good model. Unfortunately nothing of all that will be taken to heart outside the epidemiological community (if there) and climatologists will continue to treat models as date and even use them to correct actual data.

***

Somehow I missed Turchin and Spinney a decade ago. Against the prevailing postprocessualists it is possible to treat history as a science and there are rules driving human behaviour, if only you are prepared to look for them and put in the effort to distinguish the signal from noise.

***

Eberle et al. supply an important example for the above. Not that anyone is going to take any notice of observations that go against religiously held convictions. As far back as history can reach people have never done so, they are not going to start now.

Zum Anfang      Übersicht Artikel      Home & Impressum

Creative Commons Attribution-Share Alike 3.0 Unported License Viewable With Any Browser Valid HTML 4.01! Valid CSS!


1

To give a drastic example, let's assume that as of today people refrained from having sex altogether. The resulting total halt of births will occur in nine months time. No amount of data tweaking for recorded births and their slight ups and downs would give you any warning of the abrupt cessation before it actually occurs. Similarly if you change policy today no amount of nowcasting will give you the effect before the data are in.     Zurück