Urban75 Home About Offline BrixtonBuzz Contact

Coronavirus - worldwide breaking news, discussion, stats, updates and more

The one instance I saw (after the fact rather than modeling) was the virus being spread being virtually all along the air flow in the office, with very little on other floors of the building despite people touching same bannisters, lift buttons etc. That sort of modeling would seem valuable?

That doesn’t really seem like modelling
unless we are taking a really broad definition. Yes, these were real-world examples that we learned from, but they are a different thing to the detailed mathematical modelling that has let us down on some occasions.

We can start with the outdatrd influenza-based modelling (basically a model lifted from over a decade ago using a completely different virus model) that filled in a lot of unknowns with assumptions and led to the early “herd immunity” plans if you want to go into detail. That is easily Googlable. It helped to foment this whole shambles.

Models can provide useful insights but are not a replacement for empirical science, and those commissioning them are very prone to saying “just give me a number”, without any willingness to appreciate the inherent limitations.

More specifically,the particular model referenced here, quite aside from the implicit assumptions, has a whole bunch of parameters you need to input. Assuming the model is near perfect (which is ridiculously charitable), where are you getting the input parameters from?
 
Last edited:
That doesn’t really seem like modelling
unless we are taking a really broad definition. Yes, these were real-world examples that we learned from, but they are a different thing to the detailed mathematical modelling that has let us down on some occasions.

We can start with the outdatrd influenza-based modelling (basically a model lifted from over a decade ago using a completely different virus model) that filled in a lot of unknowns with assumptions and led to the early “herd immunity” plans if you want to go into detail. That is easily Googlable.

Models can provide useful insights but are not a replacement for empirical science, and those commissioning them are very prone to saying “just give me a number”, without any willingness to appreciate the inherent limitations.

Yes indeed not modeling at all, but seems the way to update/perform your (small-scale) modeling of buildings to reduce transmission.
 
Yes indeed not modeling at all, but seems the way to update/perform your (small-scale) modeling of buildings to reduce transmission.

Sure, not arguing with that. More talking about these mathematical projections that give a sense of “science” while relying on inputs that are conjecture at best.
 
More specifically,the particular model referenced here, quite aside from the implicit assumptions, has a whole bunch of parameters you need to input. Assuming the model is near perfect (which is ridiculously charitable), where are you getting the input parameters from?

Sorry, don't understand that bit. Input paramaters such as?

Was also going to say the example I gave is going to have to be built into all office (etc) air-conditioning systems. Almost needs the exhaust air outlet to be directly above every occupant, but even then you're drawing air in from around so if someone's next to you, you could still be breathing in their droplets. Not sure how displacement ventilation would affect that, too.

Eta: does look like it reduces problem rather than having air flow past lots of people, and apparently used a lot in Scandinavian countries. Wonder if that affected transmission rates within offices/industrial buildings in Sweden?

1597484459475.png
 
Last edited:
ok ta. I did look at the spreadsheet and searched 'input' but can't see examples. Are they things like number of students, spacial separation, whether they have masks, things like that?
 
ok ta. I did look at the spreadsheet and searched 'input' but can't see examples. Are they things like number of students, spacial separation, whether they have masks, things like that?

The fist point I see in the instructions starts with "specify input parameters". :confused:

But yeah, all of that and more, it seems to be based on a fluid dynamics model (would need someone closer to the subject to confirm that).
 
The one instance I saw (after the fact rather than modeling) was the virus being spread being virtually all along the air flow in the office, with very little on other floors of the building despite people touching same bannisters, lift buttons etc. That sort of modeling would seem valuable?
I think in the terms 8ball is thinking, that is a descriptive model - ie take real-world data and create a 'reconstruction scene' to show how the virus spread in that particular environment. Very useful, of course, as it shows how the virus is likely to spread in similar environments.

I think the models 8ball is thinking of that have failed to varying degrees have been the predictive ones, where a bunch of data regarding infection rates have been input into a mathematical model of spread to see what will happen in the future but without any real-world examples to draw on where what is predicted has actually happened.

The Imperial 'risk of 300k deaths' model falls into this category. And in terms of the input parameters that weren't quite right, a few I can think of are asymptomatic cases, the pattern of superspreader spread, and - probably the biggie - the assumption of 100% vulnerability of the population at the start. Just taking that last bit, there is growing evidence that not everyone is vulnerable to getting sick with C19, either cos they already have the T-Cells to fight it off or because they fight it off at the exterior of their bodies. Either they get high infection levels but fight it off without the need for new antibodies, or they just don't let it get a proper foothold in their bodies in the first place. Exactly how many such people there are makes an enormous difference to that headline '300k' figure.

I'd be interested to know how Imperial have adjusted their model over time as new information has come to light. The best models are those that update themselves in a Bayesian way - adjusting the parameters as new real-world data is fed in. One such is that of Karl Friston, of 'dark matter' fame. Friston's numbers work best when assuming high levels of non-susceptibility. Even then you have to be cautious cos that might not be the only explanation, but he predicted some of the subsequent findings of resistance which lends his model some credibility.

This sums up the approach

Intuitively, this means one is trying to optimise probabilistic beliefs—about the unknown quantities generating some data—such that the (marginal) likelihood of those data is as large as possible. The marginal likelihood2 or model evidence can always be expressed as accuracy minus complexity. This means that the best models provide an accurate account of some data as simply as possible. Therefore, the model with the highest evidence is not necessarily a description of the process generating data: rather, it is the simplest description that provides an accurate account of those data. In short, it is ‘as if’ the data were generated by this kind of model. Importantly, models with the highest evidence will generalise to new data and preclude overfitting, or overconfident predictions about outcomes that have yet to be measured. In light of this, it is imperative to select the parameters or models that maximise model evidence or variational free energy (as opposed to goodness of fit or accuracy). However, this requires the estimation of the uncertainty about model parameters and states, which is necessary to evaluate the (marginal) likelihood of the data at hand. This is why estimating uncertainty is crucial. Being able to score a model—in terms of its evidence—means that one can compare different models of the same data. This is known as Bayesian model comparison and plays an important role when testing different models or hypotheses about how the data are caused. We will see examples of this later. This aspect of dynamic causal modelling means that one does not have to commit to a particular form (i.e., parameterisation) of a model. Rather, one can explore a repertoire of plausible models and let the data decide which is the most apt.

If I understand it correctly, the way it differs from many other models is that it doesn't assume causes. Rather, it creates a range of predictions, then compares them to the real data to determine which is the most likely to have the most accurate parameters, and from there it infers causes.
 
Anyway the main reason in my book for why worst case death predictions in pandemics are unlikely to ever come fully true is that people will always take action at some point. Because even if a particular government does absolutely nothing at all to mitigate a pandemic, once deaths reach a certain level people take matters into their own hands, modify their behaviours and reduce deaths that way.

And even if we ended up in a situation where somehow the fear stemming from all the deaths didnt kick in, business as usual still doesnt happen in a pandemic because at the height of epidemics the number of staff absences from various important jobs tends to increase to a level that causes service disruption, further modifying peoples routines and behaviours.
 
Last edited:
How exactly to you determine that 300,000 would not have died if we had no lockdown and other measures? I just dont really understand how the predictions of how many would die if action wasnt taken can be tested at this stage given that we did take action.
By adjusting your assumptions based on the new knowledge and rerunning the numbers in the new model.

tbh I'm not interested in criticising or handing out blame for models that proved to be wrong. We have a lot more information now, so models should be a lot better. I'm more interested in what that means for what should be done from now onwards.
 
By adjusting your assumptions based on the new knowledge and rerunning the numbers in the new model.

tbh I'm not interested in criticising or handing out blame for models that proved to be wrong. We have a lot more information now, so models should be a lot better. I'm more interested in what that means for what should be done from now onwards.
How has it proved to be wrong?
 
How has it proved to be wrong?
As I said above, there is growing evidence that its assumption of 100% vulnerability is wrong, possibly by a very large margin. That changes the numbers hugely wrt that 300k headline figure, and it potentially changes what we ought to be doing now as it means the numbers with antibodies are only one subset of those who have been exposed to the virus.

I'm not even saying that it was wrong to act on this model's predictions back in March. Given the state of knowledge then, it was quite possibly the prudent thing to do. But you can't continue to act as if it were right as the evidence mounts that it isn't. Sadly, public political discourse in the UK rarely allows for that kind of nuance, though.
 
By adjusting your assumptions based on the new knowledge and rerunning the numbers in the new model.

tbh I'm not interested in criticising or handing out blame for models that proved to be wrong. We have a lot more information now, so models should be a lot better. I'm more interested in what that means for what should be done from now onwards.

I think there is some cross-thread leakage going on here (where I asked a question about the graph showing excess deaths) so let’s not get too derailed.

I agree with your later point about how our discourse limits discussion and action. Maybe it’s something we’ll learn from at some point.

Not sure if there has been a misinterpretation but to be clear, I think based on what evidence was coming through at the time, we should have locked down sooner.
 
Last edited:
As I said above, there is growing evidence that its assumption of 100% vulnerability is wrong, possibly by a very large margin. That changes the numbers hugely wrt that 300k headline figure, and it potentially changes what we ought to be doing now as it means the numbers with antibodies are only one subset of those who have been exposed to the virus.

I'm not even saying that it was wrong to act on this model's predictions back in March. Given the state of knowledge then, it was quite possibly the prudent thing to do. But you can't continue to act as if it were right as the evidence mounts that it isn't. Sadly, public political discourse in the UK rarely allows for that kind of nuance, though.

There was very little talk of 'vulnerability rate', rather it was the 'attack rate' and overall proportion of population that the models expected to ultimately be infected in a situation where there was zero mitigation or behavioural changes that got most of the attention. And nothing that I read back then ever came out with 100% for that. The numbers that ended up being mentioned in press conferences of the time were more like 70-80% if we did no mitigation. Lets look at what the Imperial College model said:

In the (unlikely) absence of any control measures or spontaneous changes in individual behaviour, we would expect a peak in mortality (daily deaths) to occur after approximately 3 months (Figure 1A). In such scenarios, given an estimated R0 of 2.4, we predict 81% of the GB and US populations would be infected over the course of the epidemic. Epidemic timings are approximate given the limitations of surveillance data in both countries.

(from https://www.imperial.ac.uk/media/im...-College-COVID19-NPI-modelling-16-03-2020.pdf )

Theres loads of assumptions and data that was fed into models back then that would be a bit different if we repeated the exercise now, that much is true. And it does my head in that as far as the most famous UK model was concerned, I dont think I've seen much in the way of the exercise being repeated later with a refined set of assumptions and inputs. I'm going to try to look into this myself more but it doesnt seem to be very easy to find stuff beyond the first key month. When it comes to timing errors its quite easy to see that the Imperial Colleges timing estimates were all wrong when it came to the UK, because the surveillance data they fed into it was crap and if the excuse in some SAGE minutes is anything to go by, they also failed to correct for data timing lags when feeding that data into their model. The rest of what was wrong is far less easy to establish. I would also like to update my sense of hospitalisation %ages so far in different age groups, as thats important stuff. And we know that a lot of assumptions about what proportion of cases would be asymptomatic, and the role of asymptomatic cases in transmission, were probably quite far wide of the mark. But I havent seen many attempts to calculate the implications of this. Whats happened to those various academics etc from various disciplines who produced alternative pictures from the start, which were then often used by those who wanted to argue against lockdown? Are they still out there, producing new work? Because if the things you are suggesting have really already been proven wrong, I would expect them to seize on that and run updated calculations that serve their point of view. I havent looked recently, so maybe they have, in which case I would like someone to point it out.

As with some of our previous discussions I do understand where you are coming from regarding the immunity picture. Unlike you I am utterly unclear as to what exactly has been proven on this front so far, given that there has been mitigation and behavioural changes which mean we have not had a view of an unmitigated pandemic wave. And if you want modelling to take account of what you think has been shown on this front, you need to have some alternative numbers to feed in, and a sense of why those numbers are better than the original estimates. Its no good just saying that the 100% vulnerability assumption was wrong, without providing an alternative number or at least a reasonable range. And then you actually have to go and look at exactly if and how such a number was actually used by the models in the first place.
 
Last edited:
There was very little talk of 'vulnerability rate', rather it was the 'attack rate' and overall proportion of population that the models expected to ultimately be infected in a situation where there was zero mitigation or behavioural changes that got most of the attention. And nothing that I read back then ever came out with 100% for that. The numbers that ended up being mentioned in press conferences of the time were more like 70-80% if we did no mitigation. Lets look at what the Imperial College model said:



(from https://www.imperial.ac.uk/media/im...-College-COVID19-NPI-modelling-16-03-2020.pdf )

Theres loads of assumptions and data that was fed into models back then that would be a bit different if we repeated the exercise now, that much is true. And it does my head in that as far as the most famous UK model was concerned, I dont think I've seen much in the way of the exercise being repeated later with a refined set of assumptions and inputs. I'm going to try to look into this myself more but it doesnt seem to be very easy to find stuff beyond the first key month. When it comes to timing errors its quite easy to see that the Imperial Colleges timing estimates were all wrong when it came to the UK, because the surveillance data they fed into it was crap and if the excuse in some SAGE minutes is anything to go by, they also failed to correct for data timing lags when feeding that data into their model. The rest of what was wrong is far less easy to establish. I would also like to update my sense of hospitalisation %ages so far in different age groups, as thats important stuff. And we know that a lot of assumptions about what proportion of cases would be asymptomatic, and the role of asymptomatic cases in transmission, were probably quite far wide of the mark. But I havent seen many attempts to calculate the implications of this. Whats happened to those various academics etc from various disciplines who produced alternative pictures from the start, which were then often used by those who wanted to argue against lockdown? Are they still out there, producing new work? Because if the things you are suggesting have really already been proven wrong, I would expect them to seize on that and run updated calculations that serve their point of view. I havent looked recently, so maybe they have, in which case I would like someone to point it out.

As with some of our previous discussions I do understand where you are coming from regarding the immunity picture. Unlike you I am utterly unclear as to what exactly has been proven on this front so far, given that there has been mitigation and behavioural changes which mean we have not had a view of an unmitigated pandemic wave. And if you want modelling to take account of what you think has been shown on this front, you need to have some alternative numbers to feed in, and a sense of why those numbers are better than the original estimates. Its no good just saying that the 100% vulnerability assumption was wrong, without providing an alternative number or at least a reasonable range. And then you actually have to go and look at exactly if and how such a number was actually used by the models in the first place.
I think you're talking at cross purposes.
 
I think you're talking at cross purposes.

In what way? I would address the 100% vulnerability thing more directly if there was more science with actual numbers we could actually fed into these models. As I've said to littlebabyjesus before, I am interested in the immunity picture and the possible answers for some of the large gaps in our knowledge. Its just I dont think much is proven on this front, there are just some interesting possibilities that would change some key equations if we could have great confidence that they were true. But how do we actually establish they are true? I believe someone who has more pure faith in these possibilities than me would have presented the solid evidence during these discussions if it was actually available. Until such knowledge is forthcoming, I'm going to end up at odds with that sort of stance, because it would be negligent for governments to act as if it were true, and the sort of relaxations people are angling for when they make these claims are deeply suspect and inappropriate to me at this stage of the pandemic. There will come a time when even a cautious sense of the picture will be different to how it is now, and so I hedge my bets for good reason. But I'm not going out on a limb that would leave populations vulnerable to a nasty second wave if those with an optimistic view of immunity and population susceptibility didnt turn out to have gotten it right. If they could provide a nice methodology where their ideas can be tested and where enough wiggle room is left that there is still time to put the brakes back on if reality demonstrates their assumptions were faulty, then I would try to accommodate such a plan and wouldnt necessarily be highly critical o it.

Plus we already have a situation where countries have tried relaxing stuff, its not like the level of caution is keeping everyone in a full lockdown for months longer than necessary. So its not like people calling for more relaxations and less fear are only going for the low-hanging fruit, and so I'm going to give them something of a hard time as they seem to want a degree of normality that would be absolutely absurd at the moment. And sometimes people want to have their cake and eat it - eg if you believe that draconian stuff can be avoided by doing well with things like track & trace, then you should also come to understand what that means for the range of case levels we can tolerate. In the past the message was all about keeping cases down to a level that wouldnt overwhelm the NHS, but in this current phase what they have to do is keep the number of cases, clusters etc down to a level that contact tracing systems can actually handle. And be prepared to concede that fresh restrictions are necessary for specific scenarios if your contact tracing system keeps finding problems and lots of transmission in those scenarios.

Nor should anybody expect that I am going to buy into a sales pitch where the idea that superspreaders are responsible for a huge chunk of the infections is somehow a reason to relax stuff and allow a broader range of behaviours and interactions.
 
Having said all that, I do intend to stop repeating that sort of thing every time the subject comes up, and I am currently trying to prepare myself to delve back into reading the more recent research science. Because it is entirely possible I have missed some key finding in recent months, but I need to see them in proper research paper form to get my teeth into real detail. I think it will take a few weeks before I really get going on this, and I have to say I'm not terribly optimistic about how much stuff I will find.
 
In what way? I would address the 100% vulnerability thing more directly if there was more science with actual numbers we could actually fed into these models. As I've said to littlebabyjesus before, I am interested in the immunity picture and the possible answers for some of the large gaps in our knowledge. Its just I dont think much is proven on this front, there are just some interesting possibilities that would change some key equations if we could have great confidence that they were true. But how do we actually establish they are true? I believe someone who has more pure faith in these possibilities than me would have presented the solid evidence during these discussions if it was actually available. Until such knowledge is forthcoming, I'm going to end up at odds with that sort of stance, because it would be negligent for governments to act as if it were true, and the sort of relaxations people are angling for when they make these claims are deeply suspect and inappropriate to me at this stage of the pandemic. There will come a time when even a cautious sense of the picture will be different to how it is now, and so I hedge my bets for good reason. But I'm not going out on a limb that would leave populations vulnerable to a nasty second wave if those with an optimistic view of immunity and population susceptibility didnt turn out to have gotten it right. If they could provide a nice methodology where their ideas can be tested and where enough wiggle room is left that there is still time to put the brakes back on if reality demonstrates their assumptions were faulty, then I would try to accommodate such a plan and wouldnt necessarily be highly critical o it.

Plus we already have a situation where countries have tried relaxing stuff, its not like the level of caution is keeping everyone in a full lockdown for months longer than necessary. So its not like people calling for more relaxations and less fear are only going for the low-hanging fruit, and so I'm going to give them something of a hard time as they seem to want a degree of normality that would be absolutely absurd at the moment. And sometimes people want to have their cake and eat it - eg if you believe that draconian stuff can be avoided by doing well with things like track & trace, then you should also come to understand what that means for the range of case levels we can tolerate. In the past the message was all about keeping cases down to a level that wouldnt overwhelm the NHS, but in this current phase what they have to do is keep the number of cases, clusters etc down to a level that contact tracing systems can actually handle. And be prepared to concede that fresh restrictions are necessary for specific scenarios if your contact tracing system keeps finding problems and lots of transmission in those scenarios.

Nor should anybody expect that I am going to buy into a sales pitch where the idea that superspreaders are responsible for a huge chunk of the infections is somehow a reason to relax stuff and allow a broader range of behaviours and interactions.
Well apart from anything else, I think you've taken on a huge burden of responsibility, that must be difficult and heavy to carry. These aren't decisions that you have to make. We're not going to suffer if you 'get something wrong'. It's ok not to have every datum or to have taken it all into consideration. It's not your job to keep us all safe. Go a bit easier on yourself, eh? x
 
Fear not, I dont feel much of a burden these days. There was a time burden in the first months, and I did claim that I would do everything I could not to destroy my own credibility in those first months, that I wanted to take positions that would stand the test of time. But I also told people not to put me on a pedestal, and to check my facts, and that when I look back at some messages from the early months I have some regrets and made certain mistakes.

The fact is I enjoy these discussions most of the time, right up until I hit some kind of temporary personal limit and have to take a break from certain angles. The words tend to pour out of me when I can react to other peoples angles, in a way that never seems to happen for me if I am talking about my own stance in a vacuum. And even if I bump into opinions I find frustrating, I often use them as an opportunity to get my own thoughts in order and see if I can talk any sense about the detail.

The main risk for me is that eventually I start to bore myself, which is why I'm hoping to add something new by hopefully finding some interesting science in the weeks ahead. Thanks for thinking of me anyway but dont worry, since although I developed some sense of responsibility in this pandemic there are very definite limits that stop me from getting carried away with the idea. People are generally rather clued up about the pandemic these days, if I stopped posting tomorrow then I dont think it would matter, not unless some very dangerous ideas were in the ascendence and were going largely unopposed and at the moment I dont see any signs of that.
 
hmmm - it would be nice if our flu seasons are mild.



It’s the peak of flu season in Australia. At least it should be.

Usually, flu season runs through Australia’s winter — Canada’s summer — and officials look to Australia for clues on what Canada’s upcoming flu season could look like in terms of strains, caseload and severity.

But with Australian flu numbers at their lowest in years, officials have a different takeaway this time when they look Down Under: that measures used to combat the coronavirus outbreak could be having an effect on other viruses, too.

“I think it’s telling us that influenza is preventable in the same way that COVID is preventable, to some extent, anyway,” said Dr. Lynora Saxinger, an infectious disease specialist at the University of Alberta.


flu.jpg
 

Coronavirus is a poor person's virus apparently :rolleyes:

That's a link and a half!
 
Meanwhile in Spain, where the average daily new cases have gone up from around 200 to over 4,600, thousands of anti-vaxxers and other associated nutters held a protest in Madrid against covid restrictions. :facepalm:



“Anger and sadness” is what Belen Padilla, a doctor and the vice president of the College of Doctors of Madrid”, recalls feeling when witnessing the images of thousands of anti-COVID protesters. Around 2,500 people gathered in the capital’s Plaza de Colon despite the existence of the virus, to demand their ‘freedom and rights’ as Spanish citizens.

“To deny the scientific evidence is to be an absurd person,” continues Padilla, who tells local news sources that she has not only fought to save the lives of patients infected by COVID but that she was also the victim of the virus that paralyzed the entire country, collapsed hospitals and that threatens to do it again in winter.


Chants such as, “there is no fear,” “we want to see the virus” and “we are not criminals, we want to breathe” could be heard at the protest. People have been challenging the Covid-19 security measures without keeping their distance and even hugging each other, the repetitive chant of “long live love, long live life, long live freedom” was also heard.

Faced with the obligation to wear a mask in order to stop the spread of the virus and in the midst of a wave of outbreaks, the protesters complain that this is a “ridiculous” measure and a “curtailment” of their personal freedoms. The pandemic has been described as “farce” and “lie” by several of the attendees.

 
Back
Top Bottom