Thoughts on these projections?

Geoff.D

Honorary Master
Joined
Aug 4, 2005
Messages
15,210
The reality is that many 'older' South Africans have advanced education, training, and experience, but limited job opportunities, or at least occupations in which their skills are fully utilised.

In the absence of coherent information from the government (until recently), it is natural to look at data independently and come up with some answers, or at least cast a critical eye on outlandish projections.

The first rule of any mathematical training (engineering, actuarial, etc) is to be able to make ball-park estimates to validate your calculations. If your computer program comes up with 2 + 2 = 4,000 you immediately know there is a mistake.

Real life is not always so obvious, but decades of exposure to the real world give you a sense of oversight, that those too involved in the process can miss completely.
One of the principles all these doomsayers and their models ignore time and time again is Pareto's Principle.
Look it up if you don't know what it is.
 
Last edited:

Gordon_R

Honorary Master
Joined
Jul 5, 2009
Messages
12,824
One of the principals all these doomssayers and their models ignore time and time again is Pareto's Principal.
Look it up if you don't know what it is.
I wasn't aware of that specific definition, though it is consistent with stochastic models in epidemiology. Super-spreaders are much more likely to cause infections, than standing in a supermarket queue.

Edit: Worth reading: https://en.wikipedia.org/wiki/Superspreader

Edit: It is not clear if how these models include the complex interactions in a real epidemic.

Edit: Outbreaks occur when a super-spreader enters a high risk environment such as a hospital or old age home.

I first thought it referred to a different 80:20 rule: 80% of everything is rubbish, from what you read or see on TV, to what is made in China...

P.S. We need to get you a spell checker! See: https://en.wikipedia.org/wiki/Pareto_principle
 
Last edited:

Geoff.D

Honorary Master
Joined
Aug 4, 2005
Messages
15,210
I wasn't aware of that specific definition, though it is consistent with stochastic models in epidemiology. Super-spreaders are much more likely to cause infections, than standing in a supermarket queue.

Edit: Worth reading: https://en.wikipedia.org/wiki/Superspreader

Edit: It is not clear if how these models include the complex interactions in a real epidemic.

Edit: Outbreaks occur when a super-spreader enters a high risk environment such as a hospital or old age home.

I first thought it referred to a different 80:20 rule: 80% of everything is rubbish, from what you read or see on TV, to what is made in China...

P.S. We need to get you a spell checker! See: https://en.wikipedia.org/wiki/Pareto_principle
Fixed thanks.
 

Gordon_R

Honorary Master
Joined
Jul 5, 2009
Messages
12,824
Very detailed and useful discussion on the various model projections:
A week ago many were complaining about the lack of published models by government to explain Covid-19 decision-making. On Thursday the Department of Health hosted a Zoom call in which more models were presented than could fit on a Milan catwalk.

Some models are useful. If modellers carefully explain their assumptions and present multiple scenarios, we can get a better understanding of the epidemic’s possible trajectories, or the potential of different interventions to reduce the number of infections.
But models, especially because they are surrounded by fancy equations, can give a false sense of certainty. No one truly knows how the epidemic will play out. We humans, in contrast to any other species, have an insatiable desire to know the future. But we can’t. We can only make educated guesses based on the limited information at our disposal, and when it comes to Covid-19 that information is still very limited indeed.

Some common uncertainties that stand out in the models are the rate of asymptomatic infections, how infectious SARS-CoV-2 is, how effective various interventions are, and the death rate.
 
Last edited:

Gordon_R

Honorary Master
Joined
Jul 5, 2009
Messages
12,824
A scientific paper on this topic just appeared in my newsfeed, discussing the dispersion factor and super-spreaders (mentioned by @Geoff.D).

Most of the discussion around the spread of SARS-CoV-2 has concentrated on the average number of new infections caused by each patient. Without social distancing, this reproduction number (R) is about three. But in real life, some people infect many others and others don't spread the disease at all. In fact, the latter is the norm, Lloyd-Smith says: “The consistent pattern is that the most common number is zero. Most people do not transmit.”
That's why in addition to R, scientists use a value called the dispersion factor (k), which describes how much a disease clusters. The lower k is, the more transmission comes from a small number of people. In a seminal 2005 Nature paper, Lloyd-Smith and co-authors estimated that SARS—in which superspreading played a major role—had a k of 0.16. The estimated k for MERS, which emerged in 2012, is about 0.25. In the flu pandemic of 1918, in contrast, the value was about one, indicating that clusters played less of a role.
Estimates of k for SARS-CoV-2 vary. In January, researchers at the University of Bern simulated the epidemic in China for different combinations of R and k and compared the outcomes with what had actually taken place. They concluded that k for COVID-19 is somewhat higher than for SARS and MERS. But in a March preprint, Adam Kucharski of LSHTM estimated it's only 0.1. “Probably about 10% of cases lead to 80% of the spread,” Kucharski says.
If he is right, SARS-CoV-2 needs to be introduced undetected into a new country at least four times to have an even chance of establishing itself, Kucharski says. That may explain why the virus did not take off around the world sooner after it emerged in China, and why some very early cases elsewhere—such as one in France in late December 2019, reported on 3 May—apparently failed to ignite a wider outbreak. If the Chinese epidemic was a big fire that sent sparks flying around the world, most of the sparks simply fizzled out.
This clustering is significant for models, but even more important for public health contact tracing in outbreaks:
But studying large COVID-19 clusters is harder than it seems. Many countries have not collected the kind of detailed contact tracing data needed. And the shutdowns have been so effective that they also robbed researchers of a chance to study superspreading events. (Before the shutdowns, “there was probably a 2-week window of opportunity when a lot of these data could have been collected,” Fraser says.)
 
Last edited:

Gordon_R

Honorary Master
Joined
Jul 5, 2009
Messages
12,824
Great links thanks.
This science is not entirely new, and it is not clear which (if any) of the SIR models include this dispersion factor (k), and whether they have the latest value. Anything that has such a large effect on the transmission of the virus, must have a huge effect on the model projections.
 

Geoff.D

Honorary Master
Joined
Aug 4, 2005
Messages
15,210
This science is not entirely new, and it is not clear which (if any) of the SIR models include this dispersion factor (k), and whether they have the latest value. Anything that has such a large effect on the transmission of the virus, must have a huge effect on the model projections.
During the presentations this past week, I posted a question about why Pareto's principle has not been taken into account but clearly they did not know what I was talking about because they never addressed it. I guess if I had mentioned the magic word superspreader, they may have pricked up their ears.
From what I have gone through so far, the only models that take some sort of notice of Pareto's principle is the PANDA model and the UCT guy. All the others do not on the surface anyway, take super spreading into account and almost certainly the consortium model does not.

I might have missed something though. Presentations very seldom tell you the whole story. Sometimes the verbal info during a presentation is worth more than the BS displayed on the slides.
On the surface anyway, none of the SIR or SEIR models presented allow for the k factor.
The modellers do not even bother to table their global assumptions. They do not even acknowledge that they might be subconsciously making assumptions that their audience understands all the gobbledegook they take for granted that all would understand.
The key one here is that every single person infected by Covid is equally likely to infect others.
 
Last edited:

Gordon_R

Honorary Master
Joined
Jul 5, 2009
Messages
12,824
During the presentations this part week, I posted a question about why Pareto's principle has not beenbtsken into account but clearly they did not know what I was talking about because they never addressed it. I guess if I had mentioned the magic word superspreader, they may have pricked up their ears.
From what I have gone through so far, the only models that take some sort of notice of pareto's principle is the PANDA model and the UCT guy. All the others do not on the surface anyway, take super spreading into account and almost certainly the consortium model does not.

I might have missed something though. Presentations very seldom tell you the whole story. Sometimes the verbal info during a presentation is worth more than the BS displayed on the slides.
On the surface anyway, none of the SIR or SEIR models presented do not allow for the k factor.
The modellers do not even bother to table their global assumptions. They do not even acknowledge that they might be subconsciously making assumptions that their audience understands all the gobblydegook they take for granted that all would understand.
The key one here is that every single person infected by Covid is equally likely to infect others.
Good points. I will dig around more, but it seems the most 'unrealistic' models underestimate the super-spreader component, even though that is a basic stochastic technique. Confirmation bias then leads the government to choose the model that meets their 'worst case' expectations, and ignore the others that don't.

I was thinking about super-spreaders in a different context last week. I suggested that the epidemic growth would slow once it had burned through most of the 'essential' workers such as police, who were neglecting PPE protocols, and forming large cluster outbreaks.

My thinking is still a bit muddled as to how all of these factors combine, particularly with the complex demographics of SA, but I seriously doubt if any model captures these details accurately. It is good to be prepared, but bad to be consistently alarmist.

BTW, I learned a new word for the behaviour of those such as myself who voluntarily self-isolate and social distance. The term 'hermit' is the opposite of a super-spreader!
 

Gordon_R

Honorary Master
Joined
Jul 5, 2009
Messages
12,824
The SACMC model does list a number of assumptions, but most are related to 'outputs' such as hospital stays and ICU beds: https://www.nicd.ac.za/wp-content/uploads/2020/05/SACMC_19052020_slides-for-MoH-media-briefing.pdf

One of the key assumptions listed does not differentiate duration of asymptomatic infectiousness, whereas the flowchart in slide #5 says there are separate rates in the model:
Duration of infectiousness from onset of symptoms 5 days
This is directly related to cluster outbreaks and super-spreaders, and if this is biased because sick people self-quarantine, then the projections change significantly.

The other assumption doesn't indicate any kind of differentiation or definitions, only an overall thumb-suck:
Optimistic scenario
• Lockdown reduced transmissibility by 60%
--------
Pessimistic scenario
• Lockdown reduced transmissibility by 40%
The study referred to in the AAAS link has some specific details of the formulas used to estimate k and R0: https://wellcomeopenresearch.org/articles/5-67

There is no discussion how the data would stack up in SA, with all its differences in R0 and demographics.
 
Last edited:

Cage Rattler

Well-Known Member
Joined
Nov 29, 2005
Messages
331
Research the Ethical Skeptic ... one of the extremely few out there who know how to model properly and how to verify the veracity of models...
 

Gordon_R

Honorary Master
Joined
Jul 5, 2009
Messages
12,824
Updated model projections:
Updated projections by the South African Covid-19 Modelling Consortium (SACMC) was published by the National Institute for Communicable Diseases (NICD) on Wednesday, providing the first update to projections published initially on 19 May.

"The model projects that if testing patterns remain unchanged there may be more than 408 000 detected cases by mid-July. However, prioritisation of testing may result in a reduction in detected cases to approximately 133 000," the report, dated 12 June, reads.
The latest model used data from laboratory-confirmed infections until 5 June and a detailed set of parameter estimates.

"The cumulative number of deaths by mid-July is expected to be 7 440 (3 610-14 000)."

ICU bed capacities are expected to be exceeded in the Western Cape and neighbouring Eastern Cape by the end of June, according to the updated model.
Earlier projections:
The May projections estimated that, by mid-July, the number of active symptomatic cases would be around 500 000 on an optimistic band, and 1.2 million on a pessimistic outlook, with estimated roughly 5 000 deaths by mid-July.
The SACMC is made up of key experts from several university-based institutions and is convened by Dr Harry Moultrie, a senior medical epidemiologist based at the NICD, the Modelling and Simulation Hub Africa (Masha) from the University of Cape Town, the South African DSI-NRF Centre of Excellence in Epidemiological Modelling and Analysis (Sacema) from the University of Stellenbosch, Health Economics and Epidemiology Research Office (HE2RO), which is made up of experts from the University of the Witwatersrand and Boston University School of Public Health, based in the US.
Edit: Forgot to link to the original source: https://www.nicd.ac.za/wp-content/u...port_ShortTermProjections_12062020_Final2.pdf
 
Last edited:

bromster

Expert Member
Joined
Nov 2, 2012
Messages
4,685
Who cares who has it? We need to catch up with the world and start rolling out antibody testing already.

Once you're tested, fingers crossed you shouldn't catch it again. I saw a slight concern in the form of the D614G mutation among others, but these still need to be tested for reinfection.
 

krycor

Honorary Master
Joined
Aug 4, 2005
Messages
16,393
Don’t forget, I saw mention that antibodies appear to only stay in the system for 2-3months(still capable of making it from “memory”).

So antibody passports won’t be possible and vaccine equiv is likely once they can confirm some level of inoculation like the Yellow fever (if you travel near equator you’d know about this).

It’s amazing how long it’s taking to confirm things like post exposure immunity, strain/mutation immunity etc.
 

Geoff.D

Honorary Master
Joined
Aug 4, 2005
Messages
15,210
Don’t forget, I saw mention that antibodies appear to only stay in the system for 2-3months(still capable of making it from “memory”).

So antibody passports won’t be possible and vaccine equiv is likely once they can confirm some level of inoculation like the Yellow fever (if you travel near equator you’d know about this).

It’s amazing how long it’s taking to confirm things like post exposure immunity, strain/mutation immunity etc.
Because very few countries have reached the stage where they can afford to worry about those aspects yet.
Even if you find antibodies, what will that mean?
If a person who has been infected no longer had antibodies what does that mean? Very little of this matters until we have a vaccine available for general use.
 

Gordon_R

Honorary Master
Joined
Jul 5, 2009
Messages
12,824
Projections keep changing. Flattening the curve in the W-Cape means a delayed but prolonged peak:
- The peak in the Western Cape seems to be later than was originally projected and is likely to take place from end of July to beginning of August.

- This peak is also flatter than was originally projected. This means that it will not have as many hospitalisations and deaths at the peak as originally thought.

- As a result, it is projected that 5 450 beds would be needed at the "peak" should this scenario hold. This is lower than both the original provisioning scenario from April (6 304), and the previous NCEM calibration from May (7 800).

- However, this flatter trajectory would last for longer. This means potentially more cumulative deaths of approximately 10 000 people during the pandemic, up from 9 300.

- The virus could be with us for longer than we thought, with this first peak only ending towards end of November.
 

Chris_the_Brit

High Tory
Joined
Mar 6, 2004
Messages
30,800
Do you think the last point refers to the WC or SA as a whole? I find a hard time believing a peak can be for 4 months -- with the honourable exceptions of the US and Brazil, the peaks occurred very quickly in European countries...
 

Geoff.D

Honorary Master
Joined
Aug 4, 2005
Messages
15,210
Do you think the last point refers to the WC or SA as a whole? I find a hard time believing a peak can be for 4 months -- with the honourable exceptions of the US and Brazil, the peaks occurred very quickly in European countries...
The info there comes from the EC Premier so WC only. The rest of the country is buried deep in an ANC cesspit of BS.
 

Gordon_R

Honorary Master
Joined
Jul 5, 2009
Messages
12,824
Do you think the last point refers to the WC or SA as a whole? I find a hard time believing a peak can be for 4 months -- with the honourable exceptions of the US and Brazil, the peaks occurred very quickly in European countries...
Its hard to read between the lines, particularly with 2nd and 3rd hand news reports, and it is not always clear what the terms refer to. It does not mean that the day after the peak new cases will suddenly drop to zero, since the incubation period is 5-7 days, and deaths peak about 18-25 days later.

IMO that 'peak' is a misnomer in SA, since Rt is currently around 1.4, and the epidemic will only go into decline when Rt is below 0.6 to 0.8. The chances of the necessary behavioural changes happening overnight during the current 'weak' lockdown are rather slim.

The peak in New York and Lombardy was very sharp, because it coincided with a 'hard lockdown', and significant public alarm. The situation in SA is much more like Sweden or the UK, which lasted for many months. The worst case is the US which has an unfolding wave of peaks across the country.
 
Last edited:

Geoff.D

Honorary Master
Joined
Aug 4, 2005
Messages
15,210
The mediahack dashboard calculates the Rt value for SA and WC ( not fo the other provinces, yet).
Here are the values:


1593787483046.png1593787571993.png

Based on these calcs, it is clear that WC has a good handle on the spread of the virus, has reached a value close to 1 and even for a short while below 1. It should suggest to everyone that whatever WC is doing is working fairly well. The value illustrates the point Gordon is making, that Rt must drop below 1 to ensure infections will decline, as well as the lag, and what the effects of the relaxation of the LD rules had on the WC.

The SA figures show that as a country, some of the LD measures are/were working, but that whatever SA is doing is not sufficient to drive the Rt value lower, certainly nowhere near enough to drive it down to below 1.

Now we need to hope that the other provinces, specifically GP and EC would also calculate the Rt values for comparison purposes.
 
Top