Main Coronavirus / Covid-19 Discussion Thread

[blockhseagull]

931 in UK without N I devastating and hopefully the daily figures will start dropping as lockdown takes effect.. as far as new infections go surely the rate of them has to drop significantly over next few days to prove lockdown is working

Yes these are the numbers that should slow first if lockdown is working.

The reported death toll would slow 10-14 days onwards after that because I believe that’s the average time between contracting the virus and dying.

 

[Mellotron]

Let's hope that's our peak. You'd have to hope we're getting close to it now.

 

[Is it PotG?]

Shoot 'em

https://metro.co.uk/2020/04/08/keith-norfolk-wants-sunbathers-shot-army-12525834/

 

[RossyG]

Shoot 'em

https://metro.co.uk/2020/04/08/keith-norfolk-wants-sunbathers-shot-army-12525834/

Ah, the modern media... Making "bloke down the pub reckons" nonsense into ersatz news stories.

 

[BN41Albion]

I haven't seen a comparative graph for a little while. Where are we now compared to italy/Spain two weeks ago and the likes of France?

 

[WATFORD zero]

it is cumulative. looking that the total announed series from NHS, to 27th Mar is 1695, which is ~ what the graph shows. i think they are simply representing the reports given, if there are errors in the earlier numbers they have been corrected. todays offical totals are consistent between NHS and Gov.

that is to say they probably were underreporting, which is being adjusted. the comparison is one data series made a time point to a later updated data series.

Sorry, you're right. I forget that 12 days ago, 1695 was a cumulative total and not a daily total, things have moved so fast.

We still look to be under reporting by about 30-40%, but at least it appears to be a fairly constant percentage and not increasing.

 

[Mellotron]

I haven't seen a comparative graph for a little while. Where are we now compared to italy/Spain two weeks ago and the likes of France?

 

[seagull_special]

Shoot 'em

https://metro.co.uk/2020/04/08/keith-norfolk-wants-sunbathers-shot-army-12525834/

He sounds like that Harry Enfield character who could start a fight with himself. Pretty stupid shooting people in the hand or foot as it would just add to the pressure on the NHS.

 

[:J)]

View attachment 121923

View attachment 121924

Those FT graphs are very misleading. The distance between 100 and 200 deaths is exactly the same as between 500 to 1000 (or for that matter between 1000 to 2000). The US curve should be heading practically vertically at the moment, but looks like it's only gradually rising.

 

[Machiavelli]

Regardless of whether input data and such is correct, these charts should not be made public it’s totally irresponsible to release death predictions to the public in my opinion, it is REAL people’s lives they are turning into meaningless numbers and I’ve seen dozens of people being really upset by these computer generated predictions.

In a situation so uncertain, it’s not right and these predictions have been wildly off already.

I disagree with you on this one, although there are ways in which this model is problematic (that HWT has best articulated). The reason I disagree is that there are official figures that are released on a daily basis that you could apply this reasoning to.

 

[Robdinho]

Those FT graphs are very misleading. The distance between 100 and 200 deaths is exactly the same as between 500 to 1000 (or for that matter between 1000 to 2000). The US curve should be heading practically vertically at the moment, but looks like it's only gradually rising.

It's a logarithmic scale, so a straight line means cases are going up by multiples, e.g. doubling every 2, 3 or 4 days, rather than even steps of, say, 100 per day

 

[Machiavelli]

Just a quick update. I earlier made some negative comments about Patrick Vallance that were informed in part by my recollection that a colleague who worked for him some years ago thought he was a dick. I have just checked with my colleague who reports 'no, that was XXXXX. Patrick was excellent'. We are referring here to his management skills and his ability to design and analyse an experiment.

I am happy to retract my previous skeptical comments, and happy also to add that the ringing endorsement from my colleague (who is a real expert on designing and anlysing experiments, and processing numerical biological data) means that Vallance will have the right skill set to provide the best advice.

I am very pleased about all that.

I wonder if we'll get any of the trolls on the Brexit, etc boards owning up to an error

 

[darkwolf666]

Those FT graphs are very misleading. The distance between 100 and 200 deaths is exactly the same as between 500 to 1000 (or for that matter between 1000 to 2000). The US curve should be heading practically vertically at the moment, but looks like it's only gradually rising.

I hadn't noticed that, that is a particularly shit graph scale...

 

[vegster]

Sounds like we are stuck with this for quite a while...

 

[Machiavelli]

Those FT graphs are very misleading. The distance between 100 and 200 deaths is exactly the same as between 500 to 1000 (or for that matter between 1000 to 2000). The US curve should be heading practically vertically at the moment, but looks like it's only gradually rising.

I think it's slightly more that it's easy to be misled by them, rather than they are misleading. I noticed that Johns Hopkins are also using this irregular scaling, which isn't too helpful with their figures. There must be a reason why they're using this scaling, but I can't offer it. What I can say is that these guys are at the cutting edge of mathematical modelling, so I'd trust them for using it (rather than insisting upon me being right in dismissing them as being 'misleading')

 

[:J)]

It's a logarithmic scale, so a straight line means cases are going up by multiples, e.g. doubling every 2, 3 or 4 days, rather than even steps of, say, 100 per day

Ah - thanks for the info [MENTION=2589]Robdinho[/MENTION]. Maybe the FT should flag them as such, to avoid misinterpretation by cretins like yours truly? We hear a lot about flattening the curve, but I guess these type of graphs aren't designed to show that?

 

[Billy the Fish]

Shoot 'em

https://metro.co.uk/2020/04/08/keith-norfolk-wants-sunbathers-shot-army-12525834/

Would that be added to the official death toll?

 

[CHAPPERS]

I hate these daily releases of numbers. I really don’t see what good they do for the general public.

 

[Triggaaar]

There are modellers and modellers. The above is certainly not what my modeller mate is predicting, but anything is possible when you reset the parameters. The credibility of the prediction is, however, Gaussian distributed.....

My pal's main gripe is the provenance of the data. We wrote a paper on modelling of something I work on a year or so ago and concluded the predictions had 'limited value' owing to the shitness of the wet data available. I'd suggest the COVID-19 'wet data' we are getting from around the world is a bit like shirt sizes - even my XXL Brighton shirts are all of vastly different dimension, from skin tight to down to my knees. I have a feeling that all the useful modelling will be done after the fact, later this year, when predicting how COVID-20 will manifest.

Like predicting GE outcomes, among all the predictors someone is bound to get the prediction correct, but here is a huge amount of dumb luck involved.

What I'd like is for a team who's ran the data through their new model to stop and look at the results, and ask themselves if the results are even a little realistic. If they're not, I'd like them to double check their model to work out what's gone wrong. I get the feeling that the Oxford team ended up with their preposterous conclusion (that 50% were infected over 2 weeks ago) and thought 'well that's clearly bollox, but **** it, I don't want to admit that our work was for nothing so let's send it anyway'. Cretins.