On 9/4/20 9:03 pm, David Boxall wrote:
A different perspective:
https://medium.com/@jamesjansson/covid-19-modelling-is-wrong-f7246e3dc396
It's an interesting piece of work.
But it embodies the same fundamental error as the approach that it's
criticising. He's saying:
'my modelling tool is better than your modelling tool'
But it's still 'applied science', and not 'instrumentalist' science.
Applied science is characterised by the old saw:
'when you have a hammer in your hand, everything looks like a nail'
It's driven by the tools that the individual or group has to hand, and
which they're used to working with.
Instrumentalist science instead addresses problems and seeks solutions.
So the first thing that's done is problem-analysis. With the shape of
the problem understood, tools can then be evaluated to see how well they
fit the problem.
To be fair, properly-understood real-world problems are often more
complex than any available tool can cope with.
And that's clearly the case with a pandemic - or even an epidemic within
a single, moderately homogeneous and relatively independent environment.
But to start with a preferred tool and then conceive the problem within
the constraints of that tool is to condemn the project to failure.
In this case, the Susceptible, Exposed, Infected, Recovered (SEIR)
notion is patently inadequate, e.g.
- Infected needs to be exploded into at least the levels of
{Asymptomatic, Symptomatic, Hospitalised, ICU'd, In Recovery}
- New instances need to be able to be fed into intermediate states
(e.g. ailing crew flown in from the offshore vessels)
- a heap (technical term) for the Dead seems to be missing entirely
- Recovered fails to distinguish {With Strong Antibodies, Other}
I haven't done much deterministic simulation, nor Monte Carlo
simulation, since the 1970s.
But if you don't clearly distinguish the start-point(s), the states, the
transitions, and the end-point(s), and if you don't identify the
attributes of each instance (age-range and relevant-prior-conditions,
for example), and the postulated distributions of the variables, you're
just playing games, not assisting decision-makers.
At the end, Jansson write:
> ... I would caution against relying too heavily on modelling ...
Well said, that man!
Modelling is a decision support tool, and has to be treated that way.
He continues:
> I do not fault the governments of the world acting decisively on the
basis of the modelling provided so far or the actions taken. Fast action
is an absolute necessity and interventions such as isolation can buy
society time until we determine the best path forward.
Yep, agreed with that bit too.
Obiter dicta:
I suspect that the moderately effective rather than absolute isolation,
with an emphasis on staying away from the old and infirm, was an
effective means of greatly reducing the death-toll, while allowing
steady build-up towards herd-immunity - which I suspect is likely to be
achieved before an immunisation option is available.
I also think they set the 'scare-tactics' at about the right decibels -
to 'alert' people's thoughts and actions, but keep the 'alarm' factor
under control.
And, to the credit of the people responsible for this in Australia,
they've shown some adaptation to new information and alternative modelling.
And now we want them to exercise just as effective judgement during the
peak (this week) and the wind-down phases (starting late this week, but
very gently, slowly and steadily).
Hopefully the current break-out by the Laura Norder monsters, seeking
heaps more surveillance and social control powers and resources while
they can still use the virus as a pretext, will be a short phase that
won't provide Dutton and his ilk any more scope to undermine democracy.
--
Roger Clarke mailto:roger.cla...@xamax.com.au
T: +61 2 6288 6916 http://www.xamax.com.au http://www.rogerclarke.com
Xamax Consultancy Pty Ltd 78 Sidaway St, Chapman ACT 2611 AUSTRALIA
Visiting Professor in the Faculty of Law University of N.S.W.
Visiting Professor in Computer Science Australian National University
_______________________________________________
Link mailing list
Link@mailman.anu.edu.au
http://mailman.anu.edu.au/mailman/listinfo/link
