cb25
Posts: 3,521
Likes: 2,665
|
Post by cb25 on May 10, 2020 9:14:53 GMT
Contact tracing apps may not be the panacea we wish it was. Research from John Edmunds et al (including Hannah Fry who incidentally presented Contagion! The BBC Four Pandemic): Effectiveness of isolation, testing, contact tracing and physical distancing on reducing transmission of SARS-CoV-2 in different settingswww.medrxiv.org/content/10.1101/2020.04.23.20077024v1.full.pdfIncludes the following table comparing different strategies for reducing the reproduction number... SI = Self Isolation, HH = household, CT = contact tracing. So, self isolation and contract tracing app gets us to 1.4. The only way to reduce it below 1 is to limit the number of contacts AND rely on manual contact tracing not app based. TBH I hope they're wrong. I'd be more impressed if modellers used real world data from a bunch of countries with different approaches with their models and then reported how their model's 'predictions' compared against real life.
Article in The Telegraph today about the ICL model, code for which has apparently been made public albeit after having been modified. Telegraph states "when a Swedish team applied the modified model that Imperial put into the public domain to Sweden’s strategy, it predicted 40,000 deaths by May 1 – 15 times too high."
|
|
|
Post by bernythedolt on May 10, 2020 13:51:10 GMT
cb25 Re the Telegraph piece. The ICL model is a 13 year old program, allowing ample time for its back-testing against SARS, MERS, Ebola, HIV, BSE, etc, and subsequent revisions of the model. Implies it ought to be quite refined by now. So before selecting ICL's model, would it be fair to assume HMG had monitored its predictive performance, or at least ICL's record, over the years? In other words, did the Civil Service department responsible for pandemic virus planning just panic and take the first offering that came along, or have they spent the past few years keeping abreast of which modelling teams were giving the consistently best results? After all, for decades a pandemic was always on the cards as one of the biggest threats we face, they knew it was coming one day, so I presume they were keeping on top of the tools they'd need to fight it? I hope I'm right and they chose the best model available for all the right reasons. On the other hand, perhaps I'm giving the CS too much credit and pandemic planning has been a joke. I have to admit that this is looking the more likely with every passing day...
|
|
ceejay
Posts: 971
Likes: 1,149
|
Post by ceejay on May 10, 2020 13:56:39 GMT
Late to this party but here goes - I just ticked "won't" (as opposed to can't - i have a recent phone that I usually carry everywhere).
I would probably change my mind about that if the rumoured change to a decentralised data model came about.
To be frank, I don't trust governments as far as I can throw them, certainly not with the kind of level of detail that would be available to them. They might say the data will only be handed to those that need it - really? And that it will all be deleted when it's over - double really?
In the event that possession of the app were to become a prerequisite to anything I cared about (entry to a theatre, maybe) then I'd fish out an old phone, load the app on that with an "alternate" identity and use that minimally.
If I were to be convinced that the app were a silver bullet to make the whole thing go away then I might reconsider, but it really isn't: at best it's a small contribution, with all sorts of dubious loopholes and side-effects. (E.G. - I want to get at you. So I make sure I'm in the same location as you for long enough for a contact to be registered and then self-report as being infected. Bingo - you're out of action.)
No, thanks.
|
|
cb25
Posts: 3,521
Likes: 2,665
|
Post by cb25 on May 10, 2020 14:00:22 GMT
cb25 Re the Telegraph piece. The ICL model is a 13 year old program, allowing ample time for its back-testing against SARS, MERS, Ebola, HIV, BSE, etc, and subsequent revisions of the model. Implies it ought to be quite refined by now. So before selecting ICL's model, would it be fair to assume HMG had monitored its predictive performance, or at least ICL's record, over the years? In other words, did the Civil Service department responsible for pandemic virus planning just panic and take the first offering that came along, or have they spent the past few years keeping abreast of which modelling teams were giving the consistently best results? After all, for decades a pandemic was always on the cards as one of the biggest threats we face, they knew it was coming one day, so I presume they were keeping on top of the tools they'd need to fight it? I hope I'm right and they chose the best model available for all the right reasons. On the other hand, perhaps I'm giving the CS too much credit and pandemic planning has been a joke. I have to admit that this is looking the more likely with every passing day... Like you, I would hope (but not necessarily expect) that this was the best model available, yet I've seen repeated comments about its lack of peer review (isn't that a staple of the scientific method?) and no publications I've seen of the model's predictions compared against events. I find that really curious. I've never bought into any conspiracy theory and don't with this, but I do wonder whether the lack of publicized data is down to the government wanting to go with a simple message even if it's not strictly scientific. (e.g. sitting on your backside in a park 5m from your neighbours is "bad" whereas exercising in a park 2m from your neighbours is "good". Love to see the 'science' behind that).
Disclosure: I don't consider modelling to be a science, it's just a bunch of geeks playing with numbers which may/may not tell us something about how life turns out (bit like economics).
---
ceejay with a number of publications saying Boris is going to introduce the idea of a 'threat level' for the virus, I wonder if they'll try "if you want to keep in touch with the threat level in your area - for your safety - we'd encourage you to download this second app"?
|
|
iRobot
Member of DD Central
Posts: 1,656
Likes: 2,449
|
Post by iRobot on May 10, 2020 14:12:56 GMT
|
|
cb25
Posts: 3,521
Likes: 2,665
|
Post by cb25 on May 10, 2020 14:49:30 GMT
Shame he didn't expand on this comment about the UK system "It can identify users"
On the subject of getting test results, Mail Online reports "Professor Christophe Fraser, who is advising the government on its new NHSX app, said that urgency was the key to wrestle down the spread of the outbreak. ... He told Sky News: 'A 72-hour delay really means that you're having very little impact on the epidemic. A 48-hour delay is pretty bad. 'You really need to be getting the information across in 24 hours.' "
|
|
|
Post by bernythedolt on May 10, 2020 15:44:40 GMT
Disclosure: I don't consider modelling to be a science, it's just a bunch of geeks playing with numbers which may/may not tell us something about how life turns out (bit like economics).
How about modelling the weather forecast? That "bunch of geeks playing with numbers" manages to produce a fantastically accurate prediction, many days in advance. I bet you check their work before you go out without your coat!
|
|
cb25
Posts: 3,521
Likes: 2,665
|
Post by cb25 on May 10, 2020 16:11:42 GMT
Disclosure: I don't consider modelling to be a science, it's just a bunch of geeks playing with numbers which may/may not tell us something about how life turns out (bit like economics).
How about modelling the weather forecast? That "bunch of geeks playing with numbers" manages to produce a fantastically accurate prediction, many days in advance. I bet you check their work before you go out without your coat! Still not science, but at least they publish their predictions days in advance
|
|
|
Post by bernythedolt on May 10, 2020 17:29:11 GMT
How about modelling the weather forecast? That "bunch of geeks playing with numbers" manages to produce a fantastically accurate prediction, many days in advance. I bet you check their work before you go out without your coat! Still not science, but at least they publish their predictions days in advance Our interpretations of science are obviously different, and that's fair enough. If the combination of meteorological science, mathematical science and computer science (aka the weather forecast) isn't science then I'm not sure what is. As Wikipedia puts it, "Natural science is concerned with the description, prediction, and understanding of natural phenomena based on empirical evidence from observation and experimentation" and "Weather forecasting is the application of science and technology to predict the conditions of the atmosphere for a given location and time".
|
|
cb25
Posts: 3,521
Likes: 2,665
|
Post by cb25 on May 11, 2020 9:08:54 GMT
Mail Online reports "Communities secretary Robert Jenrick revealed the app - being piloted on the Isle of Wight - may need to 'adapt' or 'move to a different model'. Fewer than 50,000 people living on the island, or 35 per cent of its population, have downloaded the app since the trial began last week."
For the technically minded, The Register has an article about the app's source code.
|
|
|
Post by Deleted on May 11, 2020 9:14:54 GMT
How about modelling the weather forecast? That "bunch of geeks playing with numbers" manages to produce a fantastically accurate prediction, many days in advance. I bet you check their work before you go out without your coat! Still not science, but at least they publish their predictions days in advance Still not science, LOL!
|
|
star dust
Member of DD Central
Posts: 2,998
Likes: 3,531
|
Post by star dust on May 21, 2020 17:27:15 GMT
Yet more evidence of the difficulties - never mind the app the contact tracers will probably 'fall over' too.
|
|
michaelc
Member of DD Central
Posts: 4,842
Likes: 2,756
|
Post by michaelc on May 21, 2020 17:40:18 GMT
Yet more evidence of the difficulties - never mind the app the contact tracers will probably 'fall over' too.
Good article and makes sense. I suppose the approach needs to be backed by legal enforcement. Not easy which is probably why it hasn't really be done so far. I do agree the Gov could have done more and I say that as broadly Boris supporting person. On the app, I'm really not worried about it being centralised or not. Is the issue that some government computer knows where I am most of the time if I carry my phone? It wouldn't bother me and I'm not sure I trust say Google any more than the UK government and they already collect such data from android phones - "anonymised" so they say.
|
|
|
Post by dan1 on May 21, 2020 21:55:55 GMT
Yet more evidence of the difficulties - never mind the app the contact tracers will probably 'fall over' too.
The Sheffield group are featuring on BBC2 Newsnight at the moment.
|
|
james100
Member of DD Central
Posts: 983
Likes: 1,190
|
Post by james100 on May 28, 2020 13:07:43 GMT
|
|