Out of date figures to justify continued restrictions *title edited by MNHQ at OP's request*

[BonnieDundee]

www.telegraph.co.uk/news/2021/06/16/latest-covid-modelling-pushed-back-june-21-based-out-of-date/

Not the first time they've done it. They did this back in November.

Is anybody else sick of them pulling this shit and then revising the figures after the decision has been voted through Parliament? We are being played for fools

Just seen a screenshot of this on Twitter. FFS Not again

It's impossible for modellers to have constant fully up to date projections.

Fitting a mathematical model to complex data often takes weeks just because of the sheer computational power required (I know, I have experience with things like this).

Modelling is always on slightly older data.

They've moved the restrictions out a month to allow all adults to have at least one jab and because the infection rates are soaring.

It also gives us more chance of unrestricted school summer holidays. i can live with their decision.

@NannyAndJohn since you know about these things could you point me in the direction of where to find the modelling and how they do the modelling?

Google isn't helping and maybe I'm being thick but can't find it on imperial college website. The only thing I found was a 17 minute long youtube video explaining 'Real Time Modelling' but I would rather read.

It's strange because I can see all this with a simple google search for the US, including what assumptions they make, how often the inputs are updated, dates when they changed the method behind the modelling and a full explanation of why.

It doesn't seem as easy to find any of this info for the UK?

It might take weeks to design a model but it clearly doesn't take weeks to run it. We now know how things might look based on data that became available a few days ago.

@NannyandJohn are you really all right with the government using false figures to get a vote through Parliament?

The PHE figures were made public 30 minutes after the modelling papers dropped, so it might be tempting to think that the new data came too late to make a difference.

Yet at Wednesday’s science and technology select committee, Dr Susan Hopkins, the deputy director of PHE’s national infection service, told MPs that the Government had known about the figures since last Friday.

It means that the Government published modelling data to bolster a delay despite already knowing it was out of date.

All the models showed a significant wave of infections in the summer, and suggested that a pause of several weeks would save thousands of lives. Yet this was based on central estimates which now cannot be correct.

Highlighting the data discrepancy at Wednesday's select committee, Aaron Bell, Conservative MP, said: “The models that we seem to be relying on to justify the extension of restrictions don't appear to be using [the PHE] numbers.

What has been late is not so much publication of data but government action. Most if not all of the restrictions that remain are because the government did not put India on the red list at the same time as Pakistan and Bangladesh.

"Yet this was based on central estimates which now cannot be correct."

They’ve done this pretty much the whole way through

My reading of the policies now is that ‘caution’ means taking the very worst possible scenario and making policy based on that. It’s almost irrelevant what the actual numbers say anymore, it’s all just based on “what ifs”

My favourite yesterday was someone saying ‘what if a new variant arises with an 80% IFR”

And so if will go on with the ‘what if’ approach....the data doesn’t matter anymore 🤷‍♀️

Apparently, people can "project" whatever nonsense they like, and shrug when it turns out to be completely wrong. I could project that 1 in 3 people who have the vaccine grow a unicorn horn, and make a pretty graph to back it up. I don't believe anything I read anymore..

Do you have a non paywalled article. I can’t see what the more up to date figures are saying from that.

As a general rule of thumb, I wouldn’t trust anything in the DT on Covid though. They have an agenda and rarely bother to hide it. On more than one occasion they’ve published an article that they must have know was complete bollocks and based on data weeks out of date.

@NannyAndJohn

It's impossible for modellers to have constant fully up to date projections.

Fitting a mathematical model to complex data often takes weeks just because of the sheer computational power required (I know, I have experience with things like this).

Weeks to fit a model? How old are your computers?

It's a joke. And SAGE models have been wrong again, and again, and again.
This is political, not scientific.

the SAGE models haven’t been wrong again and again.

People’s ability to understand them on the other hand...

@RafaIsTheKingOfClay

www.mirror.co.uk/news/uk-news/covid-models-used-delay-freedom-24336603

Quotes the telegraph but I think its in the independent too. Don't have time to search for link atm

@bumbleymummy you won't get anywhere querying Nanny's "projections" they are allowed to post any worst case scenario they find on the internet as fact, then move on to the next one as soon as it's clear it's rubbish. She's a real life chicken licken.

She's a real life chicken licken.

[quote winched]@NannyAndJohn since you know about these things could you point me in the direction of where to find the modelling and how they do the modelling?

Google isn't helping and maybe I'm being thick but can't find it on imperial college website. The only thing I found was a 17 minute long youtube video explaining 'Real Time Modelling' but I would rather read.

It's strange because I can see all this with a simple google search for the US, including what assumptions they make, how often the inputs are updated, dates when they changed the method behind the modelling and a full explanation of why.

It doesn't seem as easy to find any of this info for the UK? [/quote]
Here: assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/993427/S1289_Imperial_Roadmap_Step_4.pdf

And yes, the process of fitting the model to the data and running projections for a number of scenarios DOES take a number of weeks.

The Imperial model, for example, involves setting up a contact network structure that resembles the UK population, then uses probability to simulate the spread of disease over the network.

See here - statnet.org/nme/d3-s8-ndtv.html - for animations illustrating how this is done (I prefer the one at the bottom of the page). Note that this example contains 100 nodes whereas the Imperial network will contain 66 million nodes (one for each person in the UK), so even one single step in the model will take up a considerable amount of computing power and hence time.

The model fitting uses data starting in January 2020, so one full run of the model will need over 500 steps (one per day). See how this is getting rather time consuming? Oh, and fitting the model involves running it hundreds of thousands of times in order to obtain accurate estimates of the parameters of interest (eg. R number, secondary attack rate).

Then there are the projections, with the model having to be run thousands of times with hundreds of steps in each run for a variety of different scenarios (eg. varying degrees of vaccine efficacy, varying increases in transmission of Delta).

Contrary to belief, modelling is not just pressing a button and the results fall out immediately.

@bumbleymummy See above.

Thanks for the link @NannyAndJohn much appreciated

I managed to find a page on Scot Gov site which has every model numbered in a big list of links but still can't see the same for England. (I'm invested in the English ones because we seem to follow a pattern of generally being 'stricter' than the English regulations, dare I say, regardless of the scientific modelling).

For anyone looking for the Scottish ones:

https://www.gov.scot/collections/coronavirus-covid-19-modelling-the-epidemic/

Also NannyAndJohn I thought I was a techie until I followed that link I understand the concept of nodes and computational power at a high level (from the blockchain mostly!)... I guess I thought we'd be starting to use AI / neural networks to do this but when I look into it, the information is more about how they've just started to successfully use AI to advise on what patients will / won't benefit from being turned over in their hospital bed.

Still though, however it's done, it looks like SAGE are working with Scot Gov to provide a new updated model every week, so I'm assuming that's about the maximum these things should take for rUK?