The Unz Review • An Alternative Media Selection$
A Collection of Interesting, Important, and Controversial Perspectives Largely Excluded from the American Mainstream Media
 BlogviewJames Thompson Archive
Europe Sickens
Email This Page to Someone

 Remember My Information


Bookmark Toggle AllToCAdd to LibraryRemove from Library • B
Show CommentNext New CommentNext New ReplyRead More
ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
These buttons register your public Agreement, Disagreement, Thanks, LOL, or Troll with the selected comment. They are ONLY available to recent, frequent commenters who have saved their Name+Email using the 'Remember My Information' checkbox, and may also ONLY be used three times during any eight hour period.
Ignore Commenter Follow Commenter
Search Text Case Sensitive  Exact Words  Include Comments
List of Bookmarks

Europe is an ageing continent, with a total fertility rate of 1.6, well below the required 2.1 replacement level. The decline might be reversed, but the trend is downwards. These 747 million Europeans have a life expectancy of 79 years, and three-quarters of them live in urban settings. Of even more relevance in the time of Covid-19, the median age is 42.5 years and 20% are over 65. So, old and infertile city dwellers.

A shrinking population may not be a problem, so long as the borders can be defended. Absent the ability or the will to do so, people from other lands will come in for the spoils, as they always have done. Why not? At 34 citizens per square kilometre there is apparent space, but modern life is not directly based on agricultural productivity but rather on financial space: accumulated infrastructure, social provision, dependency ratios, and extent of immigration.

At the moment some European countries are being hit hard by the Covid-19 pandemic. Here is a snapshot of the deaths per million, as at 4th May.

Certainly, it is relevant to include important factors: when the virus took hold in a particular country; when lockdown was instituted, and how severely; the age of citizens; the density of populations; the proportion of inter-generational and single households; the sociability of citizens, and the quality of healthcare. Furthermore, countries differ in how they classify deaths. Belgium may be too inclusive, or perhaps too honest.

In terms of the percentages of citizens 60+ years, Spain and UK 25%, Belgium 26% Germany 28% and Italy 29%.

Lockdowns (of varying degrees of completeness) were begun by Italy on March 9, Spain March 14, France March 17, Belgium March 18, Germany March 22, UK March 23. 14 days is long enough for differences to accumulate, though countries had different starting points.

Here is the picture for the UK showing excess deaths.

The current rate is an incredible z= +40. This is 40 standard deviations above the mean. Signals as clear as this are rare. However well we may have imagined our systems to have operated, they are not as good as many others in Europe. In all countries, survivors are rightfully grateful to the nurses and doctors that cared for them, and to varying degrees supportive of their national policies, though understandably wishing that more had been done sooner.

Nonetheless, the United Kingdom cannot boast that it is having a good war. All the above may be relevant explanations or excuses, and it is early to come to judgment, but there is little basis for claiming that the UK response has been better than the rest of Europe.

• Category: Science • Tags: Coronavirus, Disease 
Hide 114 CommentsLeave a Comment
Commenters to FollowEndorsed Only
Trim Comments?
  1. The data in this article is quite out of date now that we’re into 05/04/2020.

  2. So, 4th of May results are out of date because it is 5th of May?

  3. TG says:

    An interesting and well thought out post.

    I do disagree that migrants are heading to Europe because of a low native fertility rate. That’s a blame-the-victim slur. The natives have a low fertility rate because of massive immigration boosting population growth and making it harder for young people to get stable jobs and raise a family.

    In the Great Depression the fertility rate of the United States plummeted, because people were afraid of having children that they could not support. This did not result in Americans going extinct, far from it: once the economy boomed, living standards shot up because economic output was not simply eaten up by an endless supply of new people, and THEN fertility rates rose, because people could see their way to reasonably supporting three kids or so.

    What, you say that America would have exited the depression even faster, if someone worried about taking care of one kid instead had six or seven? Well, many other societies tried that – and without exception, they are all still miserably poor.

    But the rich want cheap labor more than anything else, so I guess we just have to blame those lazy shiftless Europeans for not listening to their betters and behaving like good little cattle.

    • Agree: YetAnotherAnon, Gordo
    • Replies: @USA 1943
  4. dearieme says:

    Can you show us that last plot going back a few more years, please, doc? I’d like to see how one or two bad flu seasons look.

    Can I just check: that’s total deaths (excess) rather than those attributed to COVID-19, is it?

    • Replies: @James Thompson
  5. Excellent clip here with UK cardiologist Dr. Aseem Malhotra.

    And this by Dr. Paul Saladino:
    “Here’s a wake-up call for the mainstream media: We already have a powerful weapon against coronavirus, and it’s NOT a drug or vaccine, it’s changing what we put in our MOUTHS. Too bad this isn’t super sexy, requires hard work, and wouldn’t be good to all of the ultra-processed food manufacturers! There will continue to be speculation about where this virus came from for a long time but we ALREADY KNOW where this pandemic has come from: Ultraprocessed food! Agribusiness is the REAL villain here! Don’t believe me? Look at the massively increased rates of severe illness with COVID in those with metabolic dysfunction and insulin resistance. Guess what? Those are completely fixable by changing our diet. Look for this week’s podcast with UK cardiologist Aseem Malhotra in which we’ll talk all about this. Here’s the most striking point: metabolic illness can be reversed in the span of WEEKS by changing what we eat! This means that over the course of the last 6 weeks that we’ve been in this quarantine we could have protected millions against this virus and future illness if this had been the messaging in the media. BOOM! I sure hope this critical point will not be missed! That would be the greatest tragedy of them all.”

  6. The cause of this so-called pandemic and its solution couldn’t be more clear.

    Given the way corona virus is being handled, one would think we don’t realize that people die quite regularly, especially when they’re in bad condition. Now, we’re practically demanding that nobody should die from catching a microbe – that we should stay home and hold our breath until everyone is guaranteed to survive. Since when have we ever believed that? Is that how we built civilization? The civilization that we’re now destroying?

    There’s little reason for insulin-sensitive people – with healthy immune status and without metabolic disease – to stay home, wear a mask or ‘social distance’ themselves. Since they won’t be getting seriously ill, their staying home wouldn’t help ‘flatten the curve’ of sick people overburdening the healthcare system (as usual, to the expense of all of us). On the contrary, active healthy people can contribute something to the economy.

    The main benefit of herd immunity is that it will allow the country to function again. And that would be good for everyone, healthy and sickly alike. The metabolically/immunologically compromised will be vulnerable to catching the corona virus from anyone who’s contracted it and is temporarily contagious, no matter whether the carrier’s general health is good or poor. And that’s the same fix that people with poor immune function are in, always and everywhere. The answer for protecting these most vulnerable people from COVID – which is only one of the many dangers to their health that they face – can be one of two things; the best one being that they start eating right. And/or, we can build as much equipment and medical facilities, where they’re most needed, as they may require. Either of these solutions is much more viable, less disruptive and less expensive than what we’re doing now. And with either solution, healthier people would no longer be punished for possessing normal human vitality.

    While governments, health agencies and scientists take steps to upgrade the availability of care facilities, equipment and treatments, individuals should follow this
    Part A (Everyone)
    Begin a therapeutic diet to quickly upgrade and regulate the immune system. This consists of, wholly or mostly:
    Home cooked meat, oily fish, eggs (especially yolks), animal fat, bone broth, collagen or gelatin, and liver, and the elimination of corn, soy, canola, safflower, sunflower, grapeseed and rice bran oils as well as flours, sugar and prepared foods.

    Part B (those most at risk for COVID complications- individuals with high BMI or chronic health issues, or taking prescription medications, etc.)
    While following the part A protocol, take reasonable precautions to limit your exposure to possible infection from others, such as limiting time or wearing a mask when in close contact with other people.

    There is no exit strategy for this haphazard insane response. Once this over-reaction to a fairly innocuous infectious agent was accepted as being necessary, there’s no way to ever declare reversion to normalcy.
    In my opinion, rather than endlessly focussing on this not particularly interesting virus, coming up with creative signboards and banners restricting movement, wrecking people’s livelihoods and painting crosses on the pavement where one must stand, we should have been onto a more obvious problem by now. What if this HAD been a deadly pathogen? Why aren’t we prepared to quickly open special quarantine/treatment centers, disconnected from regular hospitals? And what are we going to do about it?
    This little rehearsal showed how unprepared we are should a real existential threat arise.
    But no, we must instead continue to waste our time, money and effort in playacting that this is a real biological crisis, and creating an actual breakdown in our way of life. We must continue to double down, because if we take ever more extreme action about corona, that will prove that the idiocy we’ve demonstrated thus far was necessary…..right?

  7. @dearieme

    These are excess deaths, not attributed to anything in particular.
    The European comparative data for all countries goes back to 2015.
    I had a UK graph somewhere, with much longer time span, but cannot find it at the moment.

  8. I have tracked the flu comparison elsewhere, for my small!patch. I will post an update here tomorrow. In low density locations this hasn’t been much worse than epidemic flu. In conurbations it is much more damaging.

    The UK’s problem was not following the policy it had already planned. Imperial College alarmist modelling was allowed to overturn the real science base. Also, the Conservative government’s had not replenished the PPE supplies put in place by the Coalition as they aged. So there was less PPE than planned, although the planned amount would also have fallen short. A real mass killer flu would have overwhelmed the system easily. This was a blessing in disguise; a rehearsal with live ammunition.

    • Replies: @dearieme
  9. dearieme says:
    @Philip Owen

    I recently saw the aperçu “a stitch in time costs votes”. The proposition is that voters don’t reward politicians who take sensible precautions but rather politicians who are successful at grandstanding about rescuing a country in an emergency.

    I can’t remember what inspired the writer: perhaps it was the story of California’s emergency provision, built up by the Austrian muscle man chap, being dispersed by his Democrat successor.

  10. I’ve just downloaded the R code for the ‘z-scores’ methodology, in order to go through it – because it looks to me like it’s using an inappropriate denominator.

    This is something I mentioned some time ago, when I noted that the EuroMoMo charts had far too many observations that were outside of a ±4σ range. A value 4 standard deviations above the mean should only occur once per 15,787 observations, but their charts show several such observations every year.

    Consider the standard formula for a z-score –

    It seems to me that they are using the σ of the sample mean, rather than the standard error of the sample.

    Bear in mind that the sample standard deviation (the measure they ought to be using) is

    Whereas the standard deviation of the sample mean is

    This is something that first-year students often stuff up – but usually the other way around (i.e., they use the sample standard deviation when testing things to do with the sampling distribution of the sample mean).


    So if they’re inadvertently dividing through by √N and are using 5 years of weekly data by default (which is what they say), they will get z-scores that are 16× larger than they ought to be. That doesn’t change the apparently-very-high level of recent values, but it does mean that instead of z = 40 (a 1: event), the recent score was more like z = 2.5 – something that would be expected to occur once in 81 observations.

    Anyhow… point is: they appear to be using ‘z-score’ in the same way as Viccini used ‘Inconceivable’: it does not mean what you think it means.


    Weirdly, the “What Is a z-score?” page is no longer linked from the main “How It Works” page (which can’t be access directly: only by the URL for its child items).

    Their z-score methjodology hasn’t been memoryholed as such, but it’s more like “it’s nowhere to be found for people who haven’t bookmarked it already” as opposed to “no attention is being drawn to it“.

    No matter: I have the R file now, and will look at it when I get back to my desk.

    It’s also not clear to me why they cite Farrington et al (1996) – a paper from a Psych/Criminology journal – which I just read (got it from Sci-hub which is now back online). It’s an interesting paper of itself, but it says nothing about seasonal adjustment, de-trending, or power transformation (2/3rd or otherwise). That’s assuming I’ve got the correct paper: they don’t give a full reference, which is odd.

    Farrington, D. P., Barnes, G. C., & Lambert, S. (1996). The concentration of offending in families. Legal and Criminological Psychology, 1(1), 47–63. doi:10.1111/j.2044-8333.1996.tb00306.x

    • Thanks: ic1000
    • Replies: @Kratoklastes
  11. @Kratoklastes

    Oops… I pasting in the LaTeX image ‘inline’ didn’t work.

    z = 40 (a 1: event)

    Needs to be multiline…

    z = 40 is a 1:R event, where R is given by

    That’s R weeks; divide by 52 to get the number of years, which makes it 1.41E349.

    It’s a a few hundred orders of magnitude greater than the age of the universe (1.7E10).

    So if the ‘z-score’ is a z-score, covid19 has caused an excess death rate that could be expected once in a billion billion billion billion billion billion billion billion billion billion billion billion billion billion billion billion billion billion billion billion billion billion billion billion billion billion billion billion billion billion billion billion billion billion billion billion billion billion million years.

    Check my copypasta: (10^9)^38 = 10^342, times 10^6 gets you close enough to 10^349, I reckon.

    I’m gonna say: their definition of z-score produces values that are not worth a pinch of shit.

    • Replies: @Philip Owen
  12. @Kratoclastes

    Great effort, contradict this, do an alternative z-scores calculus any-one? As important as this may be. Two other factors precede these in importance: the origin of the raw data, the data themselves, the estimation imput part, are witheld. Making garbage-in plausible. Secondly, the intentional truncated communication to the public. Advancing intentional manipulation bluntly evidenced.

  13. So, they are using the standard error of the mean, rather than the standard deviation? What a nonsense.

    My deepest apologies for not digging even deeper. I wondered how it was such a rare event, worried a little, read the details, skipped the z score explanation because what could be new about that, and then moved on elsewhere.

    Thanks very much for your deeper study.

    I presume these Euro people are getting regular salaries, and I realize I trusted them too much.

    Now, as to the value of a pinch of shit, so long as the biome is healthy, it can be ingested by a Crohn’s patient whose unhealthy biome has been preventatively killed by antibiotics and may establish a healthy digestive system.

    But, I take the point you make.

    • Replies: @dearieme
    , @ic1000
  14. dearieme says:
    @James Thompson

    How very dare you, doc? You are breaching every internet convention of hysteria, rudeness, arrogance, and ignorance.

    Moreover there are some other people here who behave in like manner. The Securitate will be after you all.

  15. ic1000 says:
    @James Thompson

    In one sense, it’s useful that the people who constructed the graph have made a hash of the maths. We have to rely instead on the Mark 1 Eyeball approach (at least until Kratoclastes provides the correct calculations for z).

    What that tells us is “compared to the visible trends of mid-2016 to early 2020, the upwards trend in excess deaths that started in March is big. Really big.”

    A key issue for z and for statistics in general is “compared to what”? The relevant question for Covid-19 is not “how does this fit into the 2016-2019 norm?” Since nations are making world-historical economic and public-health choices, a better frame would be “the experiences of modern industrialized societies.”

    The closest apples-to-apples comparison is the Spanish Flu, suggesting that the time series for which z is calculated should include 1918 and 1919. For Europe, that would also include two world wars, one depression, and the 1957 flu pandemic.

    An analogy concerns the “VIX” measure of stock volatility, which Corona-chan has raised to record highs. A z measure based on the 2009-2020 record will be quite different from one that starts a few months earlier, and thus includes the 2008 liquidity crisis.

    • Replies: @res
  16. res says:

    The closest apples-to-apples comparison is the Spanish Flu, suggesting that the time series for which z is calculated should include 1918 and 1919. For Europe, that would also include two world wars, one depression, and the 1957 flu pandemic.

    I am not sure that 1918-1919 is the closest comparison, but I agree with you that including it gives the right time frame to think about. I think this month old comment talking about the relative magnitudes of flu seasons has aged rather well. (a bit more at the link itself)

    One difficulty with this conversation is there are many different versions of a “bad flu year.”

    I would say three levels cover it pretty well.
    – severe epidemic – 1918 – 675,000 US deaths (multiply by 3 to compare with today!)
    – moderate epidemic – 1968 – 100,000 US deaths (note populations below, per capita that would be >150k now)
    – also moderate – 1957-1958 116,000 US deaths
    – typical “bad flu year” – 2018 – 61,000 US deaths
    – “typical flu year” – roughly 30-40, 000 US deaths

    CDC 2010-2019 flu statistics:

    To put the numbers above in better perspective, US populations by year.
    1918 103 million
    1958 175 million
    1968 201 million
    2020 330 million

    I would guess without extreme countermeasures COVID-19 would have been in the range between 1918 and 1968 (about an order of magnitude range in per capita US deaths). I would further guess (hopefully we will have a better idea after the fact) it would have been closer to 1968.

    IMHO a 1968 level flu would not justify the shutdowns while a 1918 level flu would (unsure about a more exact threshold, that is a large range). The uncertainty is why we have them. (note that Ron’s 1 million estimate is just over half the 1918 flu per capita US deaths)

    For even more historical data, see this later comment in that thread. Notice just how much of an outlier the Spanish Flu was!

    Trends in Recorded Influenza Mortality: United States, 1900–2004

    Their summary of results.

    Results. An overall and substantial decline in influenza-classed mortality was observed during the 20th century, from an average seasonal rate of 10.2 deaths per 100 000 population in the 1940s to 0.56 per 100 000 by the 1990s. The 1918–1919 pandemic stands out as an exceptional outlier. The 1957–1958 and 1968–1969 influenza pandemic seasons, by contrast, displayed substantial overlap in both degree of mortality and timing compared with nonpandemic seasons.

    Table 1 gives an interesting summary (by month) of 12 flu seasons from 1941-1976. WWII was a bad time for the flu in the US. There is a graphical version of the Table 1 data in Web Figure 1 from the online supplement.

    I’ll include their version of your graphic since the version starting in 1900 puts 1918 into perspective relative to the post-1930 cases. To give a little more perspective on these numbers, each 10 per 100,000 corresponds to 33,000 people in 2020 USA.

    P.S. This comment is very US-centric. Sorry about that. If anyone has similar data for the UK I would appreciate seeing it.

  17. Yes, long term data would be very instructive. I had a graph somewhere, and hope to find it one day.

    • Replies: @res
  18. In my part of Europe, COVID-19 may have peaked and by the eyesight theorem may, in terms of death toll be around twice as bad a a seasonal flu (although of course with huge accompanying containment costs). The main issue was indeed flattening the curve. The existing hospitals coped. So far as I have been able to determine, not one Dragon’s Heart hospital (Nightingale’s in England, we used up Betsi Cadwallader on a health board) bed was used. This was despite all detectable, which in light of the graph means essentially all, Care Home cases being removed to hospital.

    The curve was flattened, maybe over flattened. Why over flattened? There is no cure or yet even treatment so we are all going to get it at some point. The more the first wave gets suppressed rather than mitigated, the stronger stronger the second wave will be. Australia, NZ, South Africa, Zambia et al proved that the bug doesn’t like summer and transmits less well in low population densities. Time and capacity then exists for contact tracing. The northen hemishpere had comunity transmission from the word go.

    • Replies: @utu
  19. res says:
    @James Thompson

    Here is a short article which at least has numbers for 1918-19, 1957-58, and 1968-69:
    They give a UK fatality rate of 0.5% for 1918-19. Which, as bad as it is, is only about a third of the rate in the graph I showed above.

    I am having trouble finding much more. This book page has a graphic showing deaths/1,000 in ten English cities using a very fine time granularity from June 1, 1918 to May 31, 1918.

    At the absolute peak (middle of three, roughly November) we see a mortality rate of about 5% (!), with the earlier (July) and later (March) peaks at 1-2%.

    Pages 30 and 31 have some additional graphs which provide more context.

    • Replies: @James Thompson
  20. LondonBob says:

    Amazing they were putting covid positive patients in care homes and hospices to free up space in hospitals. This has been a catastrophic failing by the government. We would have had a better outcome had we had no lockdown and instead focussed on protecting care homes. Along with taking anything Neil Ferguson says seriously the government has failed massively. I think the closing the borders still stands, even if I think this could have been relaxed once we knew the virus was a lot less dangerous than it could have been, and that it was no way to stop it and that we were heading in to summer anyway.

    Impressed the government still isn’t reversing course yet on the lockdown.

    • Replies: @James Thompson
  21. @dearieme

    Do as I model, not as I do.

    • Replies: @Cortes
  22. @LondonBob

    The injunction “Save the NHS” was always perverse: it was supposed to save us.

    The slogan suggested that the providers were more important than patients. Yes, there was a presumption that providers had to be preserved so as to be available for patients in future, but that was silly: the worst that could happen was that for a time fewer patients would survive, but the system would still be there afterwards. The slogan should have been: Don’t all get ill at once.

    • Agree: Philip Owen
  23. @res

    Thanks. I began work at Guy’s Hospital, London, on Monday 2nd September 1968. I have no memory at all of being in the middle of a major epidemic. I can remember a week in bed with terrible flu, but that was probably in 1970 or 1971. Perhaps it was less of a big deal then. It was assumed that old people (65+) died in winter.

    Another comparator is that WWII seemed very close by. There was bomb damage everywhere. London Bridge station had corrugated tin roofs over the most damaged parts. Sections of the hospital were in ruins. My department was in a prefabricated modern building, my office in the undamaged part of a a former building. Being alarmed by flu was unlikely.

  24. Cortes says:
    @James Thompson

    I’m very good at integral and differential calculus;
    I know the scientific names of beings animalculous:
    In short, in matters vegetable, animal, and mineral,
    I am the very model of a modern Major-General”

  25. utu says:
    @Philip Owen

    “The more the first wave gets suppressed rather than mitigated, the stronger stronger the second wave will be. ”

    – I have seen this kind of thinking among flu hoaxers a lot. Knut Wittkowski is their guru. Their thinking is based on several false premises: (1) Flattened and unconstrained curves have the same area and (2) The second wave is unavoidable and (3) Inevitability of death implies that prolonging life is pointless. This kind of thinking tends to afflict unimaginative callous people.

    • Replies: @Philip Owen
  26. utu says: : Data from 24 countries. The graph illustrates: (1) At least 3 times faster rate of growth for the Covid-19 epidemic in comparison with 2018, 2017, 2016 flu season epidemics, (2) The effectiveness of countermeasures that halted the growth and reversed it in a short time. The full half width of the Covid-19 peak is two times narrower than that of the flu peaks.

    How much worse would it be w/o the countermeasures. Much worse. You can get some idea from the graph above. I looked at Sweden and using the population density as a controlling variable (population density accounts for 25% of variance) I estimated how much higher deaths totals would Europe had if it followed the Swedish approach.

    See for more details here:

    • Replies: @Stephen Dodge
    , @LondonBob
  27. @utu

    utu – pray tell, when have you ever come across as being non-callous?

    your comments were a waste of time, and unless you are a blithering idiot, which I think you are not, you do not need me to explain why.

    do better in the future.

    you are too smart to say things that are obviously lies.

    and off topic, Professor Ferguson is just as much of a hero as Don Giovanni was. Different times, different standards, but AMOR OMNIA VINCIT (that being said, he is lucky he did not try and f*ck my wife, it would have ended very badly)

    • Replies: @Daniel Chieh
    , @joe2.5
  28. LondonBob says:

    It is a lot more contagious than flu, but as only as deadly as a bad flu, so it gets compressed in to a shorter time period, which makes it a lot more dramatic and gives a false impression. As it is so contagious suppression doesn’t work once it is spreading already, indeed there is no evidence it makes much difference.

    The interview I was reading said he sees a second wave as unlikely, a Spanish flu anomaly, those countries that have had it spread can wave it goodbye.

    • Replies: @utu
  29. USA 1943 says:

    Many of The Immigrants are Low Skilled with Little or no Education, and more prone to crime Example The Rape Problem in Sweden and many other Countries that take in these Immigrants, They Will be unlikely to maintain the Countries they came to, and turn their new country into a 3rd World Crap Hole, but The People Vote for this, and they think it is better to let their country go to hell in a hand basket rather than be called Xenophobic or Racist, so they will get what they deserve. I would rather have leaders that would rather be called Xenophobic and other names Than let their Country Deteriorate, but I am a small minority apparently.

  30. utu says:

    “but as only as deadly as a bad flu”. – Stop. Go back and make an argument w/o this false premise. IFR as function of age was estimated by the team of physicists from Berkley.

    IMO they did a very good job. I have plotted their results from Table 2.

    Flu infection fatality rates are much lower.

    Population’s effective IFR depends on the population age demographics. In many European countries 20% of population is above 65 years old (Italy 23%, Germany 21.5%, Portugal 21.5%, Finland 21.2%, Sweden 19.9%, ….). The youngest European countries are Norway 16.8%, Poland 16.8%, Slovakia 15.1%, Iceland 14.4%, Ireland 13.9%). Outside of Europe many countries are much younger: Israel 11.7%, Columbia 7.6%, Vietnam 7.2%, Egypt 5.2%, Nigeria 2.8%. So one may expect that IFR in Israel will be two times lower than in Italy and in Vietnam it will be 3 times lower and in Egypt 4 times lower and in Nigeria 8 times lower than in Italy.

    Anyway, anybody can read from the graph his own risk of dying if being infected.

    • Replies: @res
  31. @Stephen Dodge

    utu is a soul of kindness

    • LOL: utu
  32. Despite Belgium having highest alleged per capita death rate for covid-19 in the world – as per chart at top above –

    (Tho Belgium gov admits a many of such deaths are merely ‘presumed’)

    Belgium is ending the lockdown this weekend

    Starting Sunday Mother’s Day 10 May, up to 4 people can visit anyone else in their home

    The next day, Monday 11 May, most shops can be open again to receive customers

    Face masks required on public transport remaining the main thing for most people

    Belgium is one of the richest countries of the world, with a full 3.5% per cent of households being millionaires, and essentially zero poverty for legal residents, given Belgium’s minimum income for everyone

    Belgium in theory doesn’t need to open not to starve … but they are opening

    Does the Belgian government, closely tied to the most inner circles of the EU, know something profounder about what covid is or isn’t?

  33. @res

    To what extent do those graphs suggest the Spanish Flu and others until the mid ’40s eliminated those who were more susceptible from the gene pool?

    Corona viruses are not HxNy and perhaps it is not productive to compare them too much with Influenza.

    • Replies: @res
  34. @utu

    YEs, I am an engineer.

  35. res says:

    Utu can be quite obnoxious (and I seem to be one of his favorite targets ; ), but he makes a valid point here. In an earlier thread I included a similar plot for some flu years in Canada:

    One issue with utu’s plot is he plots the age buckets at the lower end of the age range. This is not the convention (see plot above, they should be plotted at the middle of the age bucket) and overstates the fatality rate by age (five year offset). Though I should also note he used the more conservative table 2 IFRs (next to last column rather than the mean from the last). I am a bit confused about why he did not just use their Figure 4. Perhaps because it is harder to read?

    If you look at age 75 (70 on utu’s plot) you can see that for Covid-19 the IFR was about 2% while for the flu the IFR was about 0.01%. Notice that utu’s y-axis is in % while the plot above uses deaths per 1000 (need to divide by 10 to get %).

    So something like 200x the fatality rate at age 75 (this seems high to me, did I make a mistake somewhere?). Even if you want to fudge by allowing a factor of 2 for an especially bad flu year and another factor of 2 for cherry picking the worst hit area for COVID-19 that is still a big difference.

    P.S. One thing that intrigues me is how consistently exponential with age (linear on a semi-log plot like these) the fatality rates are.

    P.P.S. Because the IFR-age relationship is exponential rather than linear using the middle of the age bucket isn’t really correct either. Does anyone know offhand what would be correct? Perhaps I should compute the age where an exponential fatality rate at that age would equal the bucket average for different doubling times? For a simple 5 year doubling case and 3 points (1, 2, 4) you get a PWL average of 2.3 so I think to be most accurate you would plot a little to the right of the bucket center. The center is probably close enough for the doubling times we have.

    • Replies: @utu
    , @LondonBob
  36. res says:
    @Peripatetic Commenter

    That’s a good point, but I think the primary drivers of the overall decrease in flu fatalities have been things like improved hygiene and food availability along with more effective care (e.g. of resulting pneumonia).

    To address your point I would suggest looking at the pandemic years. Those were the relatively novel flu viruses so your point would be less important.

    • Replies: @Peripatetic Commenter
  37. @res

    That’s a good point, but I think the primary drivers of the overall decrease in flu fatalities have been things like improved hygiene and food availability along with more effective care (e.g. of resulting pneumonia).

    I can believe that because one of the important defenses people have against Influenza is sialic acid. Since HxNy binds to Sialic acid on cell surfaces as its method of getting into cells, humans secrete sialic acid in their mucus which filters out lots of virus particles and prevents them from getting to cells in the airways.

    Healthy individuals are more likely to have good mucus.

    However, at some time the genes coding for expressing sialic acid in mucus were probably not very prevalent in the population (and may not be prevalent in populations that do not get exposed to Influenza) and in the same manner, defenses against this novel Corona virus may not be prevalent in extant populations either …

  38. joe2.5 says:
    @Stephen Dodge

    Is this an answer to the facts and/or the calculations (all of them either verified or very plausible, although the 25% allocation to the contribution to variance may possibly have been challenged as rather arbitrary)? Otherwise said, what is your point?

  39. utu says:

    “Utu can be quite obnoxious (and I seem to be one of his favorite targets ; ), but he makes a valid point here”. – Thank you for trying to be helpful, unfortunately you did not succeed in your noble self-sacrificing full of good intentions attempt that must have cost you a lot, so I will take a dubious pleasure to correct you again, though this time I will try to be as un-obnoxious as possible.

    The plot from Canada is not of IFR or CFR, unfortunately. I think I told you that about 10 days ago: “Mortality rates in your link are not IFR, so we can’t really compare.”

    It would be good to have IFR vs. age plot for flu. The two plots show that above 55 both flu and Covid-19 mortality doubling age is around 7 years. This is interesting. Look at Fig. 2 in your Canadian link where the doubling age above 40 is about 7.7.

    “This is not the convention (see plot above, they should be plotted at the middle of the age bucket)” – There is no convention. We do not know what were buckets in Canada plot so can’t know if they used the mid point. But you are correct thinking that there is a right way of doing it. If you already know from your fisrts approximation that your function is exponential f(x)=exp(kx) then the average f_avg over bucket [x,x+∆x] equals to exp(kx)(exp(k∆x)-1)/k, so you can plot averages f_avg vs. x after dividing bucket averages from experiment data by (exp(k∆x)-1)/k or you make a similar correction to plot middle intervals then divide by (exp(k∆x/2)-exp(-k∆x/2))/k. The value of k you will get in successive approximations.

    One thing that intrigues me is how consistently exponential with age…– Yes, indeed. Once you make it to your lowest mortality age which is around 12 the curves are exponential. See Figs 1 and 2 in the Canadian paper. Why exponential? Probabilities of risk are multiplicative. But why exponential for a particular disease? Life is wearing you down in a multiplicative way? Anyway, I do not have a good answer right now.

    • Replies: @res
  40. joe2.5 says:

    Thank for these reminders. One thing is remarkable: the huge increase before and during the war years. Would be interesting to know on which criteria these deaths were attributed to flu.

    [To Peripatetic Commenter: that these are flu data (hard to know if these are all deaths caused by rhabdovirus rather than coronaviruses) doesn’t distract from its value for the purpose of epidemiologic comparison.]

  41. @Kratoklastes

    My comments on Ferguson’s model at the time are below.

    How do we stop this (non) science? Peer review will crush any alternative viewpoint. Climatology has demonstrated that 40 years of failed forecasts means nothing to the Malthusian Narrative. In fact Malthus himself has been failing since 1798.

  42. utu says:

    How wrong was Imperial College forecast? March 16 document estimated daily deaths at about 20/100,000 at the peak in the UK which would amount to 13,200/day at the peak. This forecast was for the unconstrained outbreak.

    Providing that Sweden’s indeed is an example of an unconstrained outbreak (which is not entirely true) and comparing it with Finland and Norway, two countries of similar population densities that implemented the countermeasures, we can estimate the factor by which the countermeasure reduce the daily deaths. This factor is 7.5 for Sweden and Scandinavian countries (population density corrected).

    If we apply 7.5 to 13,200 we get 1,760 deaths/day at the peak. Now we look at UK daily deaths at and we see that at maximum numbers are around 1,100 per day. It is better than factor of 2. I would say that Imperial College forecast was pretty good to be off only by 60% and I do not understand why so many people are so upset with it. It did its job. It got the UK to implement the countermeasures that saved so far 29,000*(7.5-1)=188,500 lives in the UK.

  43. res says:

    The plot from Canada is not of IFR or CFR, unfortunately. I think I told you that about 10 days ago: “Mortality rates in your link are not IFR, so we can’t really compare.”

    You are right about that (which was my mistake I was wondering about). If we are willing to make assumptions about infection rate we can make some estimates. Not very good, but do you have any better data for that?

    “This is not the convention (see plot above, they should be plotted at the middle of the age bucket)” – There is no convention. We do not know what were buckets in Canada plot so can’t know if they used the mid point.

    If you look at the Canada plot the points are at 5 year intervals on the 2.5 year points. Seems safe to guess that is midpoints of 5 year buckets. 90 is an exception. Not sure how they determined that point.

    As far as conventions and correctness, let’s just say that for this exponential case the LEAST accurate way you could plot those buckets is what you did. For the range where the doubling time was 5 years your approach results in a bit over a factor of two error.

    • Replies: @utu
    , @res
  44. LondonBob says:

    The number of dead in Britain will be similar to the bad flu seasons of winter 99 and 00, especially when you factor in the larger and older population we have now, albeit compressed to a shorter time frame.

    • Replies: @Philip Owen
  45. utu says:

    The adorable puppy is in the biting mood.

    • LOL: res
    • Replies: @res
  46. Someone has some not-so-kind words about Ferguson’s model code:

    Actually, I guess she reviewed the version Microsoft engineers spent a month cleaning up.

  47. res says:

    Seems like an odd comment to make that response to (I have made comments which deserve it more). Any substantive disagreement with what I wrote? Or should I just take the solely ad hominem response as the usual “I am annoyed and have to say something, but have nothing worthwhile to say.”

  48. @Peripatetic Commenter

    What are the odds the author claimed to be female to avoid criticism?

  49. res says:
    @Peripatetic Commenter

    That’s pretty grim, but I think Ferguson’s history is even more shocking. Why did the UK actually trust this clown’s models yet another time? Perhaps I should eliminate the ad hominem and say something like: trust the inaccurate models developed by this distinguished researcher yet another time. And why exactly is he distinguished? Is it for being consistently wrong in a direction the authorities like to hear?

    Here’s the track record from anon146:

    I thought this was a good comment from Kratoklastes.

    My disagreement with the ICL model is not its stylistic shortcomings; it’s the lack of sensitivity analysis that was performed before Ferguson started telling the political class a horror story that jibed with their desire to be seen to save the world. It’s the same shit he has pulled roughly twice a decade since the mid-90s – which indicates it wasn’t an accident.

    P.S. Here is a fairly long article on Ferguson and his model if anyone wants more background.

  50. @res

    Let us ignore any criticism of the man, although his poor track record should have been pause for thought,

    The big problem is that it seems Fauci relied on that model in dictating public policy actions to the President, and the model is invalid.

  51. @res

    The Imperial College Epidemiology Department Keystone Cops!

  52. utu says:
    @Peripatetic Commenter

    Thanks for the link. I read what Sue Denim had to say and I am not impressed. She is no a scientific programmer; she does not understand how do you simulate stochastic processes. I am not saying that the Fergusson’s program has no problems and is not poorly written and documented but her criticisms have no merit. She is not qualified. Most comments display lynch mob mentality so they can be ignored but there are several comments there that are sensible.


    “I’ve read this review.

    Firstly, I’ll stress this again, I’m not going to defend Ferguson’s model. I have not seen it. I don’t know what it’s like. I don’t know if it’s any good.

    I don’t share Ferguson’s politics, even less so those of his girlfriend.”

    “I really don’t care about Ferguson’s coding style. He is not a professional programmer, he is an epidemiological mathematical modeler. How well he does this job is the relevant question, not his coding skills. Especially not his coding skills 30 years ago judged from a modern day perspective.”


    “Absolutely spot on dr_t. You clearly know your business. I’m a bit shocked by the critique as well. As a stochastic modelling expert who has written many a ‘rat’s nest’, it is obvious to me that the seed bug which she makes a meal of is not an issue at all for this particular code as it depends on an ensemble of results. Of course, it’s nice to fix it to have reproducibiity of individual runs as that may confuse novice users, but from the perspective of the end result, it changes nothing.”


    “The stochasticity is a feature not a bug; it is used to empirically estimate uncertainty (i.e. error bars). The model *should* be run many times and the mean/average and variance of the outputs are exactly the correct approach. Highlighting the difference between two individual runs of a stochastic model is only outdone in incorrectness by highlighting a single run.”


    “Absolutely spot on dr_t. You clearly know your business. I’m a bit shocked by the critique as well. As a stochastic modelling expert who has written many a ‘rat’s nest’, it is obvious to me that the seed bug which she makes a meal of is not an issue at all for this particular code as it depends on an ensemble of results. ”

  53. @LondonBob

    In Wales it looks as if it will not be so different from 2018. Certainly not twice as many excess deaths as that particular flu.

  54. dearieme says:

    Those comments that you quote are gibberish. I used to write Monte Carlo simulations. If my programs hadn’t given me the same results for identical inputs (including the seed used for the random number generator) then I’d have known that I’d blundered. What on earth persuades anybody to argue that it doesn’t matter because a lot of outputs are going to be averaged anyway? Plain illogical. Averaging rubbish just gives you averaged rubbish.

    • Replies: @utu
  55. utu says:

    Random number generator is used many times at different stages. Each time it is called it can be reset with a different seed depending on the intermediate results or the time of the internal clock so it might not be possible to get the same result even if you run it with the same input. It is not a bug but a feature.

  56. @utu

    Dude, you have just established yourself as a buffoon.

    Please stop before you step in it some more.

    (As a programmer who knows how hard it is to make code running on multi-core processors work correctly.)

    • Replies: @utu
  57. utu says:
    @Peripatetic Commenter

    Nobody is needing your expertise on multi-core processors here. The IC program was not meant to run on multi-core. Totally irrelevant issue. It is unimportant how the program was written, whether elegant, efficient, optimal, well documented or not. People who developed it knew how to use it. It was not a commercial software that must be idiot proof. Scientific programs are judged on how they solve the mathematical problem, what kind algorithm they use and not whether there is a spaghetti code. The woman who wrote the ‘review’ is unqualified. She had to go to dictionary to find what does ‘stochastic’ mean. She worked for Google doing monkey programing not scientific programming. She could be a good programmer but she is not qualified to evaluate how does the IC program work and what it dies. I knew many good programmers who were idiots in mathematics particularly form younger generations and you would not trust them with finding roots of the trinomial equation. But they were excellent monkey programmers when you clearly explained them what to do and they would take care of all monkey business programing details.

  58. @res

    Clowns are no frauds. I mean, what Neil Ferguson does is not nice and looks a lot like being manipulative on purpose.
    So what is he? – A market character (Erich Fromm) might be an adequate description. A scientist who (habitually) trades in scientific truth for – other goods (prominence, status gains, (sex…).

    • Replies: @Philip Owen
  59. @Dieter Kief

    A veritable climatologist!

  60. res says:

    Have you done any large scale programming? If you have more than one person working on a project having understandable and documented code is important. Do you fail to understand the importance of regression tests and their apparent lack here due to the non-deterministic outputs of the program?

    For people who have done real numerical programming, one of the issues you have to deal with is doing regression tests in the face of small numerical differences. This is especially annoying when dealing with multiple platforms (probably not an issue here, but worthwhile as an example) with subtly different floating point behavior. Non-determinism is much worse.

    And let’s just revisit the conclusion of that piece. I think it bears repeating.

    My identity. Sue Denim isn’t a real person (read it out). I’ve chosen to remain anonymous partly because of the intense fighting that surrounds lockdown, but there’s also a deeper reason. This situation has come about due to rampant credentialism and I’m tired of it. As the widespread dismay by programmers demonstrates, if anyone in SAGE or the Government had shown the code to a working software engineer they happened to know, alarm bells would have been rung immediately. Instead, the Government is dominated by academics who apparently felt unable to question anything done by a fellow professor. Meanwhile, average citizens like myself are told we should never question “expertise”. Although I’ve proven my Google employment to Toby, this mentality is damaging and needs to end: please, evaluate the claims I’ve made for yourself, or ask a programmer you know and trust to evaluate them for you.

    P.S. This twitter thread from John Carmack (game developer involved in the refactoring) has some interesting comments.

    Before the GitHub team started working on the code it was a single 15k line C file that had been worked on for a decade, and some of the functions looked like they were machine translated from Fortran. There are some tropes about academic code that have grains of truth, but
    it turned out that it fared a lot better going through the gauntlet of code analysis tools I hit it with than a lot of more modern code. There is something to be said for straightforward C code. Bugs were found and fixed, but generally in paths that weren’t enabled or hit.
    Similarly, the performance scaling using OpenMP was already pretty good, and this was not the place for one of my dramatic system refactorings. Mostly, I was just a code janitor for a few weeks, but I was happy to be able to help a little.

    That was my fear — what if the code turned out to be a horror show, making all the simulations questionable? I can’t vouch for the actual algorithms, but the software engineering seems fine.

    P.P.S. This might merit a separate comment, but see this discussion

    In particular, notice weshinsley’s first comment (I believe he is part of the Imperial team, if anyone knows for sure, please respond). It is interesting that he seems to fail to understand the difference between “running something lots of times over a decade” testing and the use of regular regression tests to detect changes in behavior.

    Even more importantly, notice the exchange between gavinpotter and weshinsley. gavinpotter makes a thoughtful comment based on his modeling experience (also see his other comments). In his response weshinsley makes a very specific statement. We need to know whether or not this is true because it greatly affects this discussion. (not least his credibility)

    The code, as we run and test it, for given seeds, is deterministic when run single-threaded. Post-initialisation, it is deterministic for a given number of threads. There is one section of SetupModel.cpp function, well-discussed in other issues, which can produce statistically equivalent stochastic realisations, but not binary identical ones, using multi-threading. If determinism is required there, commenting out one line can achieve it.

    • Replies: @utu
  61. @utu

    The woman who wrote the ‘review’ is unqualified. She had to go to dictionary to find what does ‘stochastic’ mean. She worked for Google doing monkey programing not scientific programming.

    I see you are into ad-hominems.

  62. @utu

    It is unimportant how the program was written, whether elegant, efficient, optimal, well documented or not. People who developed it knew how to use it. It was not a commercial software that must be idiot proof. Scientific programs are judged on how they solve the mathematical problem, what kind algorithm they use and not whether there is a spaghetti code.

    This is a piss-poor argument but I guess I have come to expect that from you.

    That code was used to justify public policy that has cost Trillions of dollars at this stage.

    For that reason its accuracy should have been verified and it should have been reviewed by lots of software engineers with experience with large projects.

    There should never have been a hint of problems with the code, but there were lots of issues.

    • Replies: @utu
  63. utu says:
    @Peripatetic Commenter

    “That code was used to justify public policy…” – And nothing that a mob of software engineers turned critics who opined about the code has any bearing on the mathematical validity of the model that the Imperial College computer program from the Ferguson’s team simulates. Asking software engineer for an opinion on the mathematical model is like asking a typist for an opinion about the manuscript of a novel that its author typed himself. All you get is irrelevant superficial minutiae that never cuts to the crux of the matter. There is too much inflated self-importance in the software engineers crowd. Most of you guys are not much more than typists greatly enhanced with the copy paste function ability. Most of you do not understand the code you type and usually you do not understand the meaning of the output. If it the output is numerical you have no clue if the number is too large or too small. You do not know whether a negative number may mean something special or not. If the output is a curve you have no clue if the curve should be convex or concave or had an inflection point. The mathematical programmers who develop a model of a physical phenomena do not debug programs like you coders. They do it much such faster because they are aided with the meaning of the output and intermediate results. They literally know what they are doing unlike you guys. They do not need to do regression testing. The regression testing is needed because when one monkey coder is replaced with another monkey coder this is the only way to assure the continuity because the monkeys do not parse the meaning of what they are doing.

  64. utu says:

    Asking software engineer for an opinion on the mathematical model is like asking a typist for an opinion about the manuscript of a novel that its author typed himself.

    • Replies: @res
  65. Putting aside the code, there is the issue of Ferguson using a population infection rate of 50% within days for his base case when the Diamond Princess results showing 17% over three weeks was already available. At the very least, there should have been a simulation using DP stats.

    This is not hindsight. Pierre Lemouine made this observation at the time.

    • Replies: @utu
    , @res
  66. dearieme says:

    “If you have more than one person working on a project having understandable and documented code is important.” In my experience it’s important even if you (I) are the only person to use the program. Nobody’s memory is perfect so write yourself comments to tell yourself what you were doing and why; give the variables and parameters names that have mnemonic value; and ensure you make maximum efficient use of functions and subroutines – no easily tangled go-to instructions.

    Even if the mathematical model had been programmed with the highest standards of care and competence, however, it wouldn’t have meant that the model need be any bloody good. Nor that the parameter values adopted bear a close relation to reality.

    It would have been sensible to judge Ferguson’s usefulness as a modeller by his career record of wild exaggeration: foot and mouth; mad cow disease; flu outbreaks – he’s failed at all of them. I can still scarcely believe that the bugger was listened to.

    • Agree: res
  67. utu says:
    @Philip Owen

    “Ferguson using a population infection rate of 50%…” – Population infection rate is what suppose to come from the epidemic simulation model not what you assume it to be. Infection rate of 17% on the Diamond Princess is low because the epidemic on the ship was halted by putting the ship on the lockdown. Passengers were sequestered in their cabins.

    • Replies: @Philip Owen
  68. utu says:

    I just do not understand the excitement and obsession with Neil Ferguson and his simulation model of the covid epidemic. It seems that once a meme is created on the social media it cannot be extinguished with reason regardless of the subsequent facts. Here is an article that tried to rectify it already in March 26.

    No, a COVID Scientist Didn’t Walk Back His Prediction (March 26)

    “A narrative rocketed around social media earlier today: An Imperial College study said that COVID-19 could kill 500,000 Brits, but in recent testimony, Neil Ferguson, the head of the group behind the study, put the number below 20,000. Clearly the lying alarmist was walking back his ridiculous predictions!”

    “Well, no. The paper actually offered simulations of numerous scenarios. The one resulting in 500,000 deaths was one where Great Britain just carried on life as before. Other scenarios, where the country locked down whenever it was necessary to stop the disease’s spread, put death totals below 20,000.”

    But to no avail. People keep juxtaposing the 500,000 number with 31,000 total deaths in the UK (as of May 8) completely ignoring the fact that the reason the UK has ‘only’ 31,000 deaths so far is because of the social distancing, lockdowns and other countermeasures that were implemented thanks to the report and testimonies by Neil Ferguson. Neil Ferguson saved tens if not hundreds of thousands of lives so far. Too bad that the measures were not implemented 4, 6 or 8 weeks earlier.

    • Replies: @Philip Owen
  69. res says:

    I’m a software engineer (in addition to my background in the particular engineering domain involved) whose primary background is in simulation and modeling. What was your background again?

    Trust me, I have had plenty of experience with crappy software engineering practices by domain expert modelers with no software engineering background. It is no fun being in an environment where people can’t be trusted to test well enough to keep things working in a consistent reproducible fashion. I have also had experience with programmers who did not understand the technical details of the domain they were modeling. That’s not much fun either, but is a different problem.

    It is interesting that you chose an analogy where the most important thing is artistic merit. Not whether or not the model actually works consistently, correctly, understandably, and reproducibly. Doesn’t seem like the type of analogy which would be chosen by someone who actually understood what is involved. Fortunately for your analogy, authors have editors.

    • Replies: @Peripatetic Commenter
  70. res says:
    @Philip Owen

    This is not hindsight. Pierre Lemouine made this observation at the time.

    Do you have a link to this? I would like to see exactly what he said. Thanks.

    • Replies: @Philip Owen
  71. @res

    It should also be emphasized that the code is not the mathematical model.

    The code is an implementation of the mathematical model and until it is proven (with regression test suite) that the code implements the mathematical model with something like 99.999% correctness it should not be used to make public policy decisions involving trillions of dollars.

    Utu seems fundamentally unable to understand the difference between the mathematical model and the code Ferguson and others claim implements that model.

    • Replies: @res
  72. Anita says:

    So now auld Europe is returning in its ”latin” colonies?? :

    EU and Mexico announce the finalization of an updated free trade agreement

    ”Key goods exports from Mexico to the EU are transport equipment, machinery and appliances, mineral products and optical/photographic instruments. Key EU goods exports to Mexico include machinery and appliances, transport equipment, chemical products and base metals. EU services imports from Mexico are dominated by travel and transport services while Mexico’s imports from the EU consist primarily of business services, transport services, travel services, telecommunications and computer and information services. Bilateral trade in goods between the Parties is currently valued at EUR 66 billion, while bilateral trade in services is worth another EUR 19 billion.”

    Digital Trade

    ”the EU and Mexico are prohibited from requiring companies to provide access to the source code of software that they own—a major asset for companies.”

  73. @utu

    It had 4 weeks to run wild before organized quarantine took place. For 2 weeks they didn’t know it was on board. Since then lots of large group testing such as Iceland, the South Bay and NYC have come up with total infection rates of 20+/- 5%.

    • Replies: @utu
  74. @utu


    1) The whole model was biased upwards. (See other comments by me).
    2) Ferguson was pushing for lockdown and used the highest figures to panic the politicians to do things his way. (Kill all the older cattle after BSE, Kill the maximum number of hoofed animals during Foot and Mouth).
    3) Other epidemiologists, in particular Oxford, were immediately scornful about all his scenarios. The Liverpool School of Tropical Medicine was somewhat reserved too.

    • Replies: @utu
  75. utu says:
    @Philip Owen

    For unknown psychological reason you made yourself believe that the virus does not spread fast because only 17% of passengers and crew on the Diamond Princess got infected. Actually the questions the experts were asking was the opposite: why so many passengers got infected. Was it because the infection spreads so fast that they got infected prior to the lockdown when all passengers were quarantined in their cabins or because the lockdown was ineffective and passengers were infected during the lockdown from the crew delivering food and necessities or through other paths like balconies were some of them were seen w/o masks, or air conditioning system. You just got it completely wrong. I have seen you making the same point several times for over month or so. Are you incapable to rectify false beliefs?

    Jan 20 – Mr. A, a Hong Kong resident (who visited Shenzhen and Guangdong on Jan. 10) boards DP in Yokohama
    Jan 20 – DP departs Yokohama
    Jan 25 – Mr. A diembarks in Hong Kong
    Feb 1 – Hong Kong’s government announced that (COVID-19) was confirmed in Mr. A
    Feb 1 – DP stops at Okinawa and begins quarantine
    Feb 3- DP arrives at Yokohama passengers and crew are quarantined
    Feb 5 – A 14-day health observation period is set for all passengers and crew members

    On Feb 3 passengers were already confined to cabins. This means that the pre lockdown period lasted 2 weeks not 4 weeks as you falsely believe and claim.

    Mr. A had 5 days to start the infection chain that infected 712 people.

    Chronology of COVID-19 cases on the Diamond Princess cruise ship and ethical considerations: a report from Japan

    • Replies: @Philip Owen
  76. utu says:
    @Philip Owen

    “The whole model was biased upwards.” – It was because he assumed the unconstrained spread as if people were oblivious to the epidemic around them and continued going about their business as usual and this was unrealistic. People would respond to the epidemic and would begin social distancing on their own as they do in Sweden (nobody goes to movie theaters in Sweden) which is why Sweden is not totally horrible even though it horrible enough as it is 7 times worse than it could have been if the government imposed mandatory social distancing and lockdowns.

    So while Ferguson model’s prediction for the unconstrained case was unrealistic (possibly 30% too high) it accomplished what was needed in terms of the unified strict policy for the whole country to reduce fatalities by a huge factor (7-10 times). Keep in mind that the rest of European countries did exactly the same based on models from their own epidemiologists and they did not need Ferguson for it.

    Ferguson did a good job and accomplished what was needed for the UK. Too bad it was not done 4-6 weeks sooner. And if you feel left out from the decision process it is too bad but everybody else should be relieved that it was not the person who can’t parse the data from Diamond Princess making life and deaths decisions for them.

    • Replies: @res
    , @Philip Owen
  77. res says:
    @Peripatetic Commenter

    Yes. That is indeed a point worth emphasizing. One thing that seems to be missing from this discussion (is it available somewhere?) is the high level model design description (or specification). In addition to that, I assume they backtested their model against previous epidemics (especially given Ferguson’s involvement in those, was this model what he used for all of those terrible predictions as well?). Is that model backtesting available? If it is in their papers, have those results been compared to the current version of the model? BTW, those earlier examples would make great (large scale system, as opposed to component) regression tests.

    One other question. Does anyone know what sort of source control and testing regimen Imperial was using before Microsoft et al. refactored their code and put it on Github?

  78. res says:
    @Philip Owen

    Thanks! I either missed that before or did not take the time to go through it carefully (it is long). Some notes.

    The section “A brief description of the model used to perform those simulations” includes the reference chasing to find the model description and offers a summary.

    Since utu seemed to think the multi-processor capability was not important, let’s include this excerpt.

    according to supplementary notes of another paper from the same team I mentioned above, the simulations for the United States required 20,000 processor hours

    This would also explain why they did not do sensitivity testing.

    And this is also a good time to revisit the point that much of the Imperial model interpretation was based on the idea (assumption) that ICU/Ventilator capacity would be a critical factor. That does not seem to have aged well.

    P.S. His most recent post might be of interest to others here.

  79. res says:

    Keep in mind that the rest of European countries did exactly the same based on models from their own epidemiologists and they did not need Ferguson for it.

    Do you have references for those other European models?

    This 538 comparison of US models (updated today) might be of interest.

    It focuses on near future prediction though.

    If anyone is interested, after taking nearly a month off the Kinsa blog has a couple of new posts over the last few days.

    The first implies a symptoms-death period of about 18 days in Figure 1 but gives different numbers in the text under “Outbreak Peak.”

    The second describes model tweaks they have made to estimate baseline flu prevalence with social distancing measures active.

    More in their 4/10 preprint.
    Real-time detection of COVID-19 epicenters within the United States using a network of smart thermometers

    • Thanks: Johnny Rico
  80. @utu

    I have always claimed two weeks followed by two weeks ineffective internal quarantine.

    • Replies: @utu
  81. Vitamin D appears to play role in COVID-19 mortality rates

    Patients with severe deficiency are twice as likely to experience severe complications, including death

    • Replies: @res
  82. @utu

    What was needed was to concentrate scarce resources on protecting the vulnerable. Self quarantine was on its way to do a large part of that. The major problem would have been multigeneration households which are mostly found in South Asian immigrant commumities. It was already clear from Wuhan that the major risk generic factors were age, male, obesity (level 2+) and diabetes distinct from obesity. Hypertension was also a candidate at the time. There were obvious special cases like cancer patients. We won’t know who is right until next winter at the earliest. That is when the value of naturally acquired herd immunity is going to show, or not. I’m with Oxford on this. The thing spread fast and asymptomatically and most people have already had it. Whether they have antibodies remains to be seen. It may have been dealt with by inner frog immune system. In which case, lockdown was pointless even harmful as it has preserved a naive population for a ripple on the first wave and a breeding ground for the second.

    I’m in the firing line on this. According to some hazard calculator I came across, I have a 1/7 chance of dying if I catch it. Hence my current focus on raising my heart rate when I go for a run. Lockdown doesn’t resolve anything. It just defers it. Over the summer probably. Southern hemisphere and tropical countries (Brazil excepted-do they live in airconditioning? Peru is in the mountains, cool there) have not had massive community transmission.

    • Replies: @utu
  83. utu says:
    @Philip Owen

    “What was needed was to” do exactly what has been done except that 4-6 weeks sooner.

  84. res says:
    @Peripatetic Commenter

    Their method is convoluted. Here is the paper underlying that article and its abstract.
    The Possible Role of Vitamin D in Suppressing Cytokine Storm and Associated Mortality in COVID-19 Patients


    Background Large-scale data show that the mortality of COVID-19 varies dramatically across populations, although the cause of these disparities is not well understood. In this study we investigated whether severe COVID-19 is linked to Vitamin D (Vit D) deficiency. Method Daily admission, recovery and deceased rate data for patients with COVID-19 from countries with a large number of confirmed patients (Germany, South Korea (S. Korea), China (Hubei), Switzerland, Iran, UK, US, France, Spain, Italy) as of April 20, 2020 were used. The time-adjusted case mortality ratio (T-CMR) was estimated as the number of deceased patients on day N divided by the number of confirmed cases on day N-8. The adaptive average of T-CMR (A-CMR) was further calculated as a metric of COVID-19 associated mortality in different countries. Although data on Vit D level is not currently available for COVID-19 patients, we leveraged the previously established links between Vit D and C-Reactive Protein (CRP) and between CRP and severe COVID-19, respectively, to estimate the potential impact of Vit D on the reduction of severe COVID-19. Findings A link between Vit D status and COVID-19 A-CMR in the US, France, and the UK (countries with similar screening status) may exist. Combining COVID-19 patient data and prior work on Vit D and CRP levels, we show that the risk of severe COVID-19 cases among patients with severe Vit D deficiency is 17.3% while the equivalent figure for patients with normal Vit D levels is 14.6% (a reduction of 15.6%). Interpretation Given that CRP is a surrogate marker for severe COVID-19 and is associated with Vit D deficiency, our finding suggests that Vit D may reduce COVID-19 severity by suppressing cytokine storm in COVID-19 patients. Further research is needed to account for other factors through direct measurement of Vit D levels.

    The headline says “twice as likely”, but I am not seeing that borne out in the article text or the underlying paper. Did I miss something? I guess they meant this statement:

    “Our analysis shows that it might be as high as cutting the mortality rate in half,” Backman said.

    But I am not seeing that supported in the paper.

    Did you see the Indonesian paper discussed in this comment which found an odds ratio of 10 for vitamin D deficiency and dying of COVID-19?

    That seems like much more compelling evidence to me.

    • Replies: @Peripatetic Commenter
  85. @res

    Yes, I did see it.

    I guess the more papers showing low Vit D is an issue for those who contract this variant of Corona viruses the better.

    • Agree: res
    • Replies: @res
  86. res says:

    If anyone is interested, I derived a closed form solution for where the bucket should be plotted (bucket average equals value at that age) based on the doubling time (d). This helps give an idea of how serious an error is introduced by just using the bucket center (which would be x = 5 in this analysis, keep in mind that utu plotted his values at the equivalent of x = 0). Including my intermediate steps in case anyone wants to check (or just understand better), but the last link has the answer along with a plot.

    First, compute the integral of the exponential 2 ^ (x/d) from 0 to 10 (i.e. over a single bucket). d is the doubling time.

    Then divide that by 10 for the average, equate to the point value at 2 ^ (x/d), and solve for x.*%28%28-1+%2B+1024%5E%281%2Fd%29%29+d%29%2Flog%282%29+%3D+2+%5E+%28x%2Fd%29

    That gives the closed form solution which we then plot for doubling times from 1 to 20.

    For a five year doubling time the bucket average should be plotted at about 5.57 years. So about a half year right of the bucket midpoint. Lines up nicely with my simple analysis earlier.

    P.S. utu, this is how you use math to solve a problem. Not just throwing equations around and acting like you did something useful as you did in comment 40. Also notice how my simple numerical analysis in the comment 36 PPS was more useful for giving an estimate than your comment 40 reply was.

  87. The data prove that sultry CoronaChan is a h8er, and she discriminates against people who vote for Andrew Cuomo and Bill de Blasio.

  88. utu says:
    @Philip Owen

    Philip Owen : It had 4 weeks to run wild before organized quarantine took place.

    Philip Owen : I have always claimed two weeks followed by two weeks ineffective internal quarantine.

    Not just ignoramus but also a liar.

    • Replies: @Philip Owen
  89. Have there been any papers that demonstrate that the current round of antibody tests can distinguish antibodies against any random Corona virus and the SARS-COV-2 virus?

  90. @utu

    The two statements are the same. “organized quarantine” is saying that earlier attempts at quarantine were disorganized and ineffective.

    • Replies: @utu
  91. dearieme says:

    Back to the program. “Although there are many items in the new version that could be criticized, in this summary we emphasize the widespread use of global variables”. Dear God, that’s pathetic. Indeed that would have been thought pathetic in the late sixties and a sure sign of intellectual laziness.

    Warning: very long and wonkishThe Imperial College modelers released the source code a couple of days ago to the model…

    Posted by Scarlett Strong on Friday, May 8, 2020

  92. @dearieme

    in this summary we emphasize the widespread use of global variables and in particular c-language structs to model data types.

    Say what?

    While I agree with the issue about global variables … c-language structs are very little different that C++ structs or classes (except they do not have methods) or Java classes (minus the methods) etc.

    Now, I noticed that the reviewer mentioned classes, but I suspect those who worked on the code did not have the time to turn everything into classes.

    It would be interesting to see the original code.

    • Replies: @res
  93. res says:

    Thanks. And my head was already sufficiently exploded by the original code being a single 15,000 line C file.

    I’m surprised the widespread use of global variables persisted into the refactored code. Too difficult to remove?

    The “zombie assumption” discussion is interesting. Anyone have any thoughts on that?

    This seems like the most damning point. And it is not about the shortcomings of the code.

    6. Out of Sample Performance

    The acid test for any model is whether it can predict successfully out of sample. There has been no evidence offered of the model’s ability to forecast. However, we do have a natural experiment to fall back on.

    In “Intervention strategies against COVID-19 and their estimated impact on Swedish healthcare capacity,” the authors of that study re-implemented the Imperial College model and applied it to Sweden. An examination of the model documentation and the model source code (also written in c and on github), shows it is the same model. The Swedish version of the model made clear short term predictions of the number of fatalities that would occur in Sweden if it followed its announced laissez-faire policy of social distancing and how much those fatalities would be reduced if various other policies were followed that are similar to those employed in the US and the UK. In chart A of figure 4 in the paper, by May 9 when this review was written, the model predicts about 100K deaths if Sweden followed its announced policy and about 25K deaths if it adopted the most stringent social distancing policies. On May 9, the actual number of fatalities in Sweden was 3,175 deaths. Thus, the model massively over-predicted fatalities. The zombie assumption is the likely problem.

    Here is a link to the study mentioned. Also see the comments there.

    Their code is available at

    P.S. What has me confused right now is how John Carmack gave a positive review of the Imperial code.

    P.P.S. She made a shorter followup post. Her comments there are worth reading, but here is the post itself.

    Seth Corey requested a shorter bullet point version of the model review of the Imperial College model, so here goes:

    • basic structure of the model has been documented for previous uses of the model, but the model has not been fully documented for its coronavirus use
    • we don’t have original model code so we can’t tell whether what has been documented was actually implemented in the model code
    • model code we do have is poorly written in a way that is prone to bugs
    – code manipulates billions of data points but all data points are fully accessible everywhere in the program rather than being protected from inadvertent access
    – code logic is convoluted and difficult to follow
    • model assumes that people take no action to protect themselves from an epidemic in the absence of government policies and thus significantly overestimates fatalities, cases, hospitalization, etc.
    • no tests were implemented to check the correctness of the model code
    • there were no benchmark models employed to sanity check the output
    • implementation of the model for Sweden massively overestimated fatalities

    • Replies: @James Thompson
    , @utu
  94. res says:
    @Peripatetic Commenter

    While I agree with the issue about global variables … c-language structs are very little different that C++ structs or classes (except they do not have methods) or Java classes (minus the methods) etc.

    They also lack support for public/protected/private declarations. And combined with the lack of methods that means the underlying data is accessible for change by anyone with a pointer. And since it sounds like many of these variables are global…

    There are a variety of ways to accomplish things like this in C:

    But I would be surprised if the Imperial programmers did so in their single 15,000 line C file.

    • Replies: @Peripatetic Commenter
  95. @res

    Yes. I know all these. I write lots of code in C. We use a variety of tools to check for problems including things like valgrind although it cannot really test for aliasing or for code accessing stuff it should not through errant pointers.

    There are lots of large programs written in C. The Linux kernel for example … and people know how to get it right.

    Secondly, writing in C++ or Java is not a panacea. The programmers have to know how to use things like private and protected and all the class mechanisms.

    And he/they still made lots of other mistakes, like the one you pointed out above about not testing with out-of-sample data etc.

    • Agree: res
  96. utu says:
    @Philip Owen

    No, you said “It had 4 weeks to run wild before organized quarantine took place.”

    • Replies: @Philip Owen
  97. dearieme says:

    The shortcomings of the programming (as we used to say: now everybody seems to say “coding”) are a worry for two reasons. (i) It presumably introduced computational errors – as the Edinburgh study showed. (ii) It demonstrates that the workers were not conscientious, intelligent, and competent.

    That latter point, in turn, explains why the model – however well or badly coded – is lousy. For example, it doesn’t represent the natural instinct of people to try to protect themselves from infection.

    Perhaps I’m being unfair to them : maybe it wasn’t incompetence and intellectual laziness that led them to write and use such a biased model. Maybe they did it for political reasons, or even just for the unscrupulous careerist motives that play such a large part in modern mass-society “science”. Oh the pity!

    • Replies: @Philip Owen
    , @Dieter Kief
  98. @utu

    Are you deliberately being dim? Disorganized ineffecrive quarantine is not quarantine.

    • Replies: @utu
  99. @dearieme

    Generating alarm brings fame, influence and money for more research. Tinge that influence with politcal motives to add a bit more incentive. There we have the driver for climatologists everywhere.

  100. @dearieme


    I agree. I’d add a few ands though and delete a few of your ors.

    Then I’d sum up like this: The market-character (Erich Fromm), which is embodied in a prototypical manner by Neil Ferguson, is leaning towards the Anywheres (David Goodhart) these days – as does – deep down in his multicultural heart – Boris Johnson.

    If I look at this problem the other way round, I find two prominent politicians in Europe, who were by and large for the Taiwan- etc. strategy, including closed borders and the canceling of flights from China and other hotspots of the infection: Matteo Salvini in Italy and Alice Weidel in Germany (both from late February on – and both members of parties of Goodharts Somewheres).

    Btw. Alice Wiedel spoke in great detail of her proposals for a (temporary) shut-down in march the 4th in the German Bundesparliament in Berlin – and was laughed at, ridiculed, and scoffed at by all other party-members in the Bundestag. Main argument of all of them (which caused lots of applause): To effectively fight CO-19, we need not isolate ourselves. To the contrary: Only international cooperation can by any means help us out and therefor the cancelation of international flights and the closing of our borders would be counterproductive… – That caused big waves of applause and high spirits in the aisles of the Anywhere-parties (85% of the German Bundestag). They triumped over tough Alice Weidel.

    A few weeks later an article by Angela Merkel appeared in the Frankfurter Allgemeine Zeitung. – And Merkel emphasized again and again in it – how important it is in this crisis, not to cancel “all our international connections”. And she goes on and on how interconnected the modern world is and how science and progress depend on openness and on internationality and unhindered exchange cross open borders, etc. pp.

    When Czechia decided to close its borders and make the use of masks in public mandatory, the public opinion was very quick to decide in Germany, that this was just another sign of the strength of the dangerous populism, which is about to conquer and destroy now middle European democracies – be it Czechia, Slovakia, Poland or – God beware! – Hungary…


    There are a few exceptions to this pattern – : – established rather liberal politicians who did not succumb to the Somewhere/Anywhere dichotomy in Europe: Namely chancellor Stefan Kurz in Austria, who did close Austrian borders rather quickly, and Mette Frederiksen in Denmark, who did the same.

    Then there is the very interesting and successful example of Albania (11 deaths per Million now, even though they have very close ties to Northern Italy) – the Albanians did notice what had taken place in Singapore and Taiwan and acted accordingly very quick! – When asked how that was possible, they could laugh – and show their smartphones…

    • Thanks: dearieme
  101. @dearieme

    Appropriately long, highly informative.
    What stands out for me, as an old-time amateur academic programmer, is that this program has some simple errors, most egregiously the lack of CHECK SUM in which totals are printed out at various stages so that you can check that each step is making sense. I still use that in Excel spreadsheets to check that I am reading the correct variables, and applying updating currency rates correctly.
    Later in life I was an advisor to proper programmers, and they spent all their time on standards, designing appropriate “architectures” and checking ever step again and again, until everything worked as required.
    The other observation is one I wish I had made, because it is purely psychological: the “do nothing” option did not mean “the Govt does nothing” but “neither Govt nor any citizen does anything”. Yes, the test case should be to note what many people do during epidemics: retreat to their homes, check with friends whether they have “the flu” and ask them not to visit if they have; and if possible move to more secluded and out of the way places.
    So, thanks very much for posting this.

    • Replies: @utu
  102. @res

    Brief, and highly damning.

  103. utu says:
    @Philip Owen

    How did we get here? It was your very embarrassing statement “there is the issue of Ferguson using a population infection rate of 50% within days for his base case when the Diamond Princess results showing 17%” which demonstrated your confusion: (1) the infection rate is model’s output not its input and (2) your false belief that 17% infection rate on the DP sets an upper limit for the infection prevalence for Covid-19. The DP was quarantined and the virus spread that reached 17% prevalence was constrained. For this reason the DP ‘experiment’ tells us nothing about R0 or the herd immunity threshold of the Covid-19 in general. There is one thing however which the DP data were very useful for. The IFR for older age bracket was determined pretty accurately as the denominator was accurately known. In fact this IFR was used to scale Italian data by the team of physics from Berkley to derive age depended IFR in Italy based on excess mortality data:

    Total COVID-19 Mortality in Italy: Excess Mortality and Age Dependence through Time-Series Analysis

  104. utu says:

    {“global variables” =>”Dear God, that’s pathetic.”} =>”Dear God, that’s pathetic.”

  105. utu says:

    “Out of Sample Performance” – Bad language, bad analogy, bad metaphor. Sure giveaway of a rant; not an objective analysis. The concept of “out of sample performance” does not apply in this case. Whoever said it must have been thinking about fitting or training an empirical model on one sample and verifying it on another sample. This is not how it works in this case. The epidemic progression model is not data driven. It is not a bunch of generic, usually linear, equations like in AI for which you need to find coefficients using known empirical outputs that the model is suppose to simulate. The parameters in the epidemic progression model have specific physical meanings. Some of them have to be guesses based from experience and some can be measured from the disease behavior in other places. They are not tweaks. The parameters are disease specific and population specific. The case of Sweden was not an out of sample performance test. Whoever used the IC software on Sweden must have used wrong parameters. Sweden has 10 times lower population density than the UK. Sweden has the highest single resident households in Europe. The parameters that were used for the UK should not have been used for Sweden.

    • Replies: @res
  106. I have been looking for articles on antibodies against COVId-19 and whether they can be distinguished from antibodies against other Corona Viruses.

    This seems to have some useful info:

    While there are multiple antigens to choose from, Nucleoprotein is a strong contender. The Nucleoprotein is a vital structural protein with the primary function of forming a complex with viral RNA to mediate packaging and replication. While not as immunodominant as Spike, it is highly immunogenic and is profusely over-expressed during infection [18].

    The challenge in using Nucleoprotein, however, is its similarity between coronavirus strains. Unlike Spike, Nucleoprotein shows less genetic variation – especially between SARS-CoV-2 and the genetically distinct, 2002 SARS Coronavirus. The problem here, is that antibodies to SARS-CoV Nucleoprotein have the potential to cross-react with SARS-CoV-2 Nucleoprotein (i.e. the same antibodies are able to bind Nucleoprotein from both coronaviruses). Indeed, studies have shown the coronavirus Nucleoprotein to be broadly cross-reactive [19][20].

  107. res says:

    Did that actually mean anything? It is both entertaining and educational to watch you comment on something you know little about yet have strong opinions on. And the hands wave furiously on…

    • Replies: @Peripatetic Commenter
  108. @res

    It is so amusing to see you spank his so badly.

    out of sample data Noun: data you did not use to build your model so you can be sure you are not wanking!

  109. utu says:
    @James Thompson

    “..most egregiously the lack of CHECK SUM in which totals are printed out at various stages so that you can check that each step is making sense.” – I don’t believe you unless you were an accountant. There is no room or need for “CHECK SUM” in most of mathematical programming. CHECK SUM of what? This only makes sense in transmission of numeric data or text. You can look up “Numerical Recipes in C: The Art of Scientific Computing” to see that no algorithm uses checksums and checksums are explicitly mentioned only when the sequence of bit is transmitted on page 896.

    • Replies: @Peripatetic Commenter
  110. @utu

    I don’t believe you unless you were an accountant. There is no room or need for “CHECK SUM” in most of mathematical programming. CHECK SUM of what? This only makes sense in transmission of numeric data or text.

    Oh boy, you seem to be very ignorant.

    These days ‘checksums’ are known as ‘hashes’ and there are many different types of hashes that can be used.

    We use them everywhere in computing. If mathematical programming does not use them they are very backward, but I suspect it is you who is backward.

    • Replies: @utu
  111. utu says:
    @Peripatetic Commenter

    “We use them everywhere in computing” – Perhaps, probably because your kind of ‘computing’ needs it, but as I said it you do not need it in most of scientific programming applications.

  112. res says:
    @Peripatetic Commenter

    In that spirit (I don’t think this is as compelling as the Indonesian paper), here is another article (and paper) about vitamin D.

    This is intriguing.

    This study shows that, counter intuitively, countries at lower latitude and typically sunny countries, such as Spain and Northern Italy, had low concentrations of vitamin D and high rates of vitamin D deficiency. These countries also experienced the highest infection and death rates in Europe.

    The northern latitude countries of Norway, Finland, and Sweden, have higher vitamin D levels despite less UVB sunlight exposure, because supplementation and fortification of foods is more common. These Nordic countries have lower COVID-19 infection and death rates. The correlation between low vitamin D levels and death from COVID-19 is statistically significant.

    And the paper:

    I found it odd that they mentioned Portugal as an outlier (see Figure 1) when Scotland and Sweden look like the biggest outliers to me.

Current Commenter

Leave a Reply -

 Remember My InformationWhy?
 Email Replies to my Comment
Submitted comments have been licensed to The Unz Review and may be republished elsewhere at the sole discretion of the latter
Commenting Disabled While in Translation Mode
Subscribe to This Comment Thread via RSS Subscribe to All James Thompson Comments via RSS