Coronavirus

[Old Grumpy]

indeed. Success!

That's one thing less to worry about.

Thanks, all.

OG

[Old Grumpy]

Excellent. You might need to run it one more time empty, as sometimes the tissues break up, and you get lots of bits. On the other hand it might all be clear. Either way it'll probably be OK, except that if there are bits of tissue, they'll get all over the clothes and be a bit of a nuisance.

Yep! I don't know if some tissues are paper blended wit something else, they don't all seem to fall apart. I do have to remind other members of the household from time to time not to leave them in their pockets

[Bryn]

Yep! I don't know if some tissues are paper blended wit something else, they don't all seem to fall apart. I do have to remind other members of the household from time to time not to leave them in their pockets

And to check the filter every few months.

[Anastasius]

...
The trouble with IT is that it's not always possible to know, or control what people do....

Many are too busy trying to order stuff from Wickes so they can spend the next three weeks on paid holiday doing DIY

[Dave2002]

And lo: https://www.theguardian.com/science/...s-of-modelling

Unsurprisingly, it appears to have been taken out of context/seriously overinterpreted in parts of the media.

If matching a model to the data using lots of computer time there's not so much wrong with that. If that is what was done, it would only give some parameter estimates, but there would be no reason to choose any particular outcome over any other. Choosing the most optimistic outcome without justification isn't a good strategy, neither is choosing the worst. Splitting the difference and going half way - not really any better. So basically it's not a very good predictive model - though it might become one if it started to track future trends well, but as it says in the Guardian article - it may get people thinking.

The one thing which may not have changed between the worst case and the best case - assuming that the modelling has any validity at all - is the death rate on infection, which remains at 0.14 - not very encouraging.

Also, does the model which might have been used for the curve matching also take into account possible changes of strategy, which may or may not have effects? As yet it seems unlikely that the model "knows" about the changes in human behaviour which hopefully should modify the outputs.

[Anastasius]

Our local medical centre now allows patients over 75 to request repeat prescriptions by phone, giving 5 days' notice, and the medicines will be delivered. (For some reason, this is not mentioned on its website).

That's a bit antiquated. I do it online.

[cloughie]

That's a bit antiquated. I do it online.

Eigh, lay off the oldies - if it works don’t fix it!

[Anastasius]

Eigh, lay off the oldies - if it works don’t fix it!

Ahem...I'm also an oldie !

[LMcD]

That's a bit antiquated. I do it online.

The person for whom I ascertained this information is recovering from an unsuccessful operation intended to save the sight in one eye following a detached retina, which might make online ordering a bit tricky for her were she online (which she's not).
As it happens, I've been using a very efficient online pharmacy for several months.
I just happen to like helping other people .

[LeMartinPecheur]

I've been watching with concern government statistics about number of cases in each area of local gov't since spotting that Hampshire, my home county and one where I have friends and family, is always top of the list. So to get a better basis for comparison, I've used today's official stats to work out the number of cases per head of population (actually, per notional million heads of pop. to keep the numbers sensible) for loads of southern counties (mainly). Results:

Dorset 35.10
Somerset 37.56
E Sussex 37.84
Kent 45.89
(N Yorks 48.78)
Cornwall 53.00
Devon 60.38
W Sussex 60.54
Wilts 72.20
Surrey 92.44
Oxfordshire 100.30
Hampshire 150.49 (some jump!)

The above are based on local government areas so that the Hants figures don't include Portsmouth and Southampton. So I did those separately:

Portsmouth 134.88
Southampton 94.87

Bad certainly, esp. Pompey, but not as bad as mainly rural Hants. And I'd have guessed that Surrey, pretty built-up and closer to London, would be higher than Hants but it's not. How come? One would hope that HMG or local politicians are asking questions. OK, the extreme uncertainty/ variability inherent in measuring numbers of cases may be some of the reason but it'd be nice to know. (Perhaps Hampshire have simply invested in more testing kits?)

Any suggestions?

[Simon B]

If matching a model to the data using lots of computer time there's not so much wrong with that. If that is what was done, it would only give some parameter estimates, but there would be no reason to choose any particular outcome over any other. Choosing the most optimistic outcome without justification isn't a good strategy, neither is choosing the worst. Splitting the difference and going half way - not really any better.

Agreed. I think the paper is showing the probability density functions (mostly Gaussian) of the a posteriori estimates of different scenarios which are solutions to the model and cluster around 3 sets. So they've not just picked one extreme outlier solution for its exceptional notability.

However, to start trying to critique at that level requires an exact understanding of what they've done so is the realm of peer review.

So basically it's not a very good predictive model - though it might become one if it started to track future trends well, but as it says in the Guardian article - it may get people thinking.

Any model where lots of key parameters are unknowns is likely to be poor for prediction. Though it may even turn out to track the future trends well, despite the values of the driving parameters all being wrong. Such is the case in completely unrelated fields anyway (bitter experience)...

The one thing which may not have changed between the worst case and the best case - assuming that the modelling has any validity at all - is the death rate on infection, which remains at 0.14 - not very encouraging.

Correct me if I'm wrong, but in the paper, 0.14 is the conditional probability of death once inside the set with "severe disease". So the broader case mortality rate is much lower.

Also, does the model which might have been used for the curve matching also take into account possible changes of strategy, which may or may not have effects?

They expressly excluded data beyond the point where external measures (distancing etc) may have changed the underlying system - because the model assumes it is time-invariant.

Not disagreeing with you on anything. Just trying to make sense of it as one whose method of trying to understand the world tends to work this way.

[kernelbogey]

I've been watching with concern government statistics about number of cases in each area of local gov't since spotting that Hampshire, my home county and one where I have friends and family, is always top of the list.....Any suggestions?

A Hampshire resident writes:
Aaaargh!
First and so far only thought on this: we have a high number of commuters to London from this county. I did this for about 20 years (1975-2000). Then, trains were packed with commuters from Southampton, Winchester, Basingstoke; and they included also folk who commuted from the Isle of Wight, Bournemouth and Poole.

London has the highest density of cases now, so I wonder whether commuters spread the virus outward before we were alert to it, in the (still unidentified) period when those infected may be both asymptomatic and contagious.

[LeMartinPecheur]

A Hampshire resident writes:
Aaaargh!
First and so far only thought on this: we have a high number of commuters to London from this county. I did this for about 20 years (1975-2000). Then, trains were packed with commuters from Southampton, Winchester, Basingstoke; and they included also folk who commuted from the Isle of Wight, Bournemouth and Poole.

London has the highest density of cases now, so I wonder whether commuters spread the virus outward before we were alert to it, in the (still unidentified) period when those infected may be both asymptomatic and contagious.

kernel: I fully accept your point about packed commuter trains but wouldn't we still expect Surrey to be worse? And I'd have thought Kent's figures would be far higher on this basis.

[teamsaint]

From just over the Hants / Wilts border, which you can see from just up the road....

A friend from the Waterside tells me that there was/ is a big cluster around Basingstoke, which you’d have to guess is a commuter based thing, as KB suggests. It would have to be a really big cluster to distort the figure that far, but with the cities not included in the figures, Basingstoke would he a pretty big proportion of what is left.

[LHC]

A Hampshire resident writes:
Aaaargh!
First and so far only thought on this: we have a high number of commuters to London from this county. I did this for about 20 years (1975-2000). Then, trains were packed with commuters from Southampton, Winchester, Basingstoke; and they included also folk who commuted from the Isle of Wight, Bournemouth and Poole.

London has the highest density of cases now, so I wonder whether commuters spread the virus outward before we were alert to it, in the (still unidentified) period when those infected may be both asymptomatic and contagious.

It’s worth remembering that the first cases came in with people returning from holidays and conferences abroad. The areas they returned to, which are likely to include richer areas and the commuter belt, may still have larger numbers of infections than might otherwise be expected simply because they were ‘ground zero’ for infections in the UK.