Coronavirus (2021) thread

[BerkshireBlue]

I had a quick look at that paper and what stood out to me was: "CovidSim contains over 900 input parameters".

That for me is a model killer. If you can't model something with less than 10 variables then throwing 900 at it screams of overfitting to the extreme. Not only will your model be slow to react but your predictions will be worthless as well

 

[grunge]

This time round, on average around 7% lower than the "optimistic" model. Still that's better than the 17% or so they were wrong by against the "optimistic" model against the Alpha variant.
They only undershot on the AZ 1 dose for hospitalisation figure.

which model are you looking at? there were 3 of them?

 

[BerkshireBlue]

which model are you looking at? there were 3 of them?

Well there was this crap one from Warwick. They're currently 35% out from their best guess and that's going to get worse and worse over the next few days. I reckon they'll be over 60% off when the 21st June hospitalisation numbers are released. They predicted 530 admission on the 21st, my model said 280 I reckon it's going actually be sub 250 at this rate.

 

[grunge]

That for me is a model killer. If you can't model something with less than 10 variables then throwing 900 at it screams of overfitting to the extreme. Not only will your model be slow to react but your predictions will be worthless as well

Yep, the scope for cascading errors with 900 variables is obscene. Unless a lot of those are properly encapsulated out into sub classes/coroutines

example: if the model has a ( random guess from top of my head ) social distancing routine. that provides a few variables that feeds into the model but could be a whole sub routine to create that variable that needs 50/60 variables in its own right to run the code that calculates the single output,

 

[grunge]

View attachment 19894

Well there was this crap one from Warwick. They're current 35% out from their best guess and that's going to get worse and worse over the next few days. I reckon they'll be over 60% off when the 21st June hospitalisation numbers are released. They predicted 530 admission on the 21st, my model said 280 I reckon it's going actually be sub 250 at this rate.

Ta, I was looking at this one.

https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/984521/S1227_SPI-M-O_Summary_of_further_modelling_of_easing_restrictions_Roadmap_Step_3__2_.pdf

 

[BLUEMATT23]

Australia have strict quarantine for 14 days, don’t they?

This proves that you cannot stop a virus by shutting borders. I accept that you can slow down the number of clusters, but this idea of keeping it out is (and always has been) nonsense.

It's absolutely crazy, Australia and NZ are targeting "zero covid", will never ever happen. Some of the clowns in charge here also are seeking this but it is unrealistic and pointless.

 

[BlueAnorak]

Australia have strict quarantine for 14 days, don’t they?

This proves that you cannot stop a virus by shutting borders. I accept that you can slow down the number of clusters, but this idea of keeping it out is (and always has been) nonsense.

Yes but with the new mRNA and Vector vaccines you can now lock down your borders to give yourself time to develop a vaccine and deploy it to limit serious illness and death.

 

[PannickAtTheDisco]

Prof Ferguson "almost certainly probably 50% less infectious".

This is what we're dealing with.

 

[BlueAnorak]

which model are you looking at? there were 3 of them?

Warwick and LSHTM.
It gobsmacked me that the press has not actually compared these models to the actual data in the PHE report.
Should I have time this month, I will knock up a spreadsheet.
I really shouldn't be gobsmacked, our press across the board is absilutely useless. All they could be bothered to do was knock up a few stories about the actual vaccine data being better than the reports.

 

[Gremlin]

Yes but with the new mRNA and Vector vaccines you can now lock down your borders to give yourself time to develop a vaccine and deploy it to limit serious illness and death.

Except they haven't done that effectively down there. My brother and his family live in Aus and they are exasperated by the slow pace of vaccine roll-out there. And all the while they risk an outbreak that undoes everything before they've got their vaccine protection in place.