OT: Review of Ferguson/Imperial model

[newshound]

Shows you can't be too careful. The fake news people are getting much smarter. Not sure what the answer is, though.

[Andy Burns]

I don't know if it runs multi-threaded or not, if so that could complicate things, probably have to re-seed the random numbers every time it forks/spawns something?

[Pancho]

Very easy. It is often the case that the default constructor of a Pseudo Random Number Generator is without a seed parameter and uses something like the time the code starts as a seed instead. A lazy coder will just use the default and only worry about testability later.

On multithread code race conditions are dependent on the cpu scheduling, and are effectively non deterministic. Even running single threaded you can get race conditions from things like timed events.

I would need to look at the code myself to know.

For running, averaging is fine, it might even converge. The problem is testing, if you're not testing you are on shaky ground.

[Vir Campestris]

The experiment I would like to see is to take a completely typical bit of the UK, and test everyone possible for the presence of antibodies.

That would give us an idea the true infection rate in the country.

Without that we don't know the death rate.

If it turns out the 30,000 people that have died are the result of half the population being infected then we'd probably be better off letting another 30k dies, instead of the excess deaths caused by the breakdown of the NHS (it's not overloaded, but that's because lots of people who should be using it aren't) and plain poverty.

If on the other hand those 30,000 are the deaths that have resulted from

0.1% of the population being infected - be very scared.

The trouble is we've got a PM who studied classics, a health secretary who studied economics, and a leader of the opposition who is a lawyer.

None of them would know a good model or experiment if it bit them on the arse.

Andy

[Andy Burns]

But which bit is typical?

[Jake56]

We don?t have a reliable antibody test yet.

Only in that particular area.

That particular number isn't very useful.

Its very unlikely that half the population has been infected.

Can't see that being sellable to the voters.

That?s wrong.

That?s not going to happen because a lot more than that have tested positive.

Irrelevant to what their scientific advisors have studied.

The reason for the lockdown was Ferguson's modelling.

They don?t need to.

[Andy Burns]

Rather than sample one area, we could sample all of a particular type of worker, e.g. all firemen or all postal workers whose forename begins with A, F or S etc chosen to get a reasonable sample size.

[Clive Page]

I'd like to see that too.

Actually I wonder if they are going to use the Isle of Wight population for that as well as to test out their new-fangled tracing App?

[Clive Page]

But the PM's leading advisor (Cummings) is another arts man, who pretends to understand science. That may be the worst kind of all.

[Jake56]

He's not a SCIENTIFIC advisor.

[The Natural Philosopher]

Once again we see that we are in a situation where science, the doyen of prediction, can't help.

All it can do is say 'lockdown will help' with some certainty.

[Martin Brown]

We once met the guy who designed that auction and watched in amazement as he got a bunch of testosterone fuelled senior managers to buy a £1 coin for £2. I dropped out at my top bid of 99p. Several went past £1 but two just kept on going! It illustrated his point very well.

When people get really involved in competitive bidding logic goes out of the window and a winner takes all mentality sets in at silly prices.

[Chris J Dixon]

Indeed! I have sold some fairly ordinary stuff on ebay that, with postage, came quite close to the cost of a new item.

Chris

[Andrew]

Neil Fergusons original code has not been released, from what I gather. It seems that Microsoft, or some other outfit have been through it and done a 'redacting' job.

[charles]

A small local charity with which I was involved, benefited from this once. We were going to get half the proceeds on an auction of promises at a nearby golf club. - we got £26,000. (and that was 24 years ago)

[Tim Streater]

Or you can have a raffle and simply sell both halves of the tickets. No need even to have real prizes then.

[Martin Brown]

Not a redacting job so much as making it suitable for their compiler. Prehistoric legacy C doesn't sit well with modern compilers.

[tim...]

but it wasn't a 1 pound coin was it?

There are millions of other 1 pound coins if you don't win this one

the spectrum auction was a one time opportunity to buy the product,

you either overpaid or didn't have a future business model

they aren't comparable, just because walking away with a ruined current business is better than keeping a loss making one

[Andy Burns]

I think the point being that if you can persuade someone to over-pay for a £1 coin with a known value, it should be a piece of piss to get them to overpay for spectrum with perceived extra value

[Vir Campestris]

I'm not sure where the hoax bit comes in.

It was a 15000 line single file.

That's really all I need to know to be certain that this code does not follow the requirements of modern software engineering. Requirements that have evolved from decades of bad experiences.

Andy