[chbot] Chchrobotics Digest, Vol 152, Issue 6

Bevin Brett bevin_brett at hotmail.com
Sun May 10 08:22:20 BST 2020


> I believe your modelling may not be elaborate enough, and that any crude
home approach is not very meaningful. Are details of the modelling
method of those uni researchers available who advised the government?

> Your model ending with an infection rate of 97-99% does not seem
plausible to me. Herd immunity should kick in before then.

Yes, their models are available, and they are vary widely.

I made no attempt to get the numbers to model specifically NZ society and Covid-19.  Instead I am interested in the stability of the modelling - how does changes in the inputs affect the outputs.  What I was really investigating was how plausible are the models.

It turns out herd immunity is VERY sensitive to the infectious behavior of the disease and society.   As I change the probability of one contact passing on an infection from 3% to 20% the number of people protected by herd immunity plummets.

If this carries over to other models and to real life, you can see why getting the correct value P for Covid-19 could be important, and - given the variations in society and the presentation of the disease, probably impossible.

P        %unexposed when all over
0.03   99%
0.06   29%
0.09   13%
0.12   7%
0.15   4%
0.18   2%
0.21   1%

So the answer is - they are not much better than a wild ass guesses, which is why the different models have been making widely different projections, and why the same model being given new data is also making very different projections.

You might want to follow Carl Bergstrom on Twitter to see some of the discussion.

/Bevin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ourshack.com/pipermail/chchrobotics/attachments/20200510/fb180193/attachment-0001.html>


More information about the Chchrobotics mailing list