Well, yes, yes they are

JohnM says:
March 19 2021 at 7:05 pm
Scotland has more people testing positive, per 100,000, than the rest of the UK, @73.4 per 100,000


Richard Murphy says:
March 19 2021 at 7:33 pm
But the elderly are getting double doses

Which would seem to indicate that first doses first has something to recommend it then….

14 thoughts on “Well, yes, yes they are”

  1. But don’t forget that at these levels of results, the false positives will probably outnumber the true positives. Don’t fall into the same trap as the rest of the media – base rate neglect

  2. “But don’t forget that at these levels of results, the false positives will probably outnumber the true positives. Don’t fall into the same trap as the rest of the media – base rate neglect”
    While false positives and false negatives are indeed a thing in any testing regime it should certainly be possible to use statistical methods to identify and correct the errors.
    What is the false positive rate (and implied base rate) for the UK testing regime?

  3. Bloke in North Dorset

    The FP for LFDs was thought to be around 0.3%, but recent schools testing is coming out with lower numbers testing positive, so its not clear.


    How much of this is driven by the tests not being carried out correctly or even carried out is open to speculation. I saw a Tweet from one guy who is using them to carry out experiments with his children. Every day they test something different, never the child. There’s also the issue of kids doing self administered testing, how many 13 year olds are going to be sticking cotton buds up their noses until it hurts?

    Its a different issue with PCR tests.

    The test itself can pick up tiny fragments of dead virus and this is often referred to as a false positive because the tested person isn’t infectious but may still have lasting symptoms.

    This is not to be conflated with someone who has never had the virus and still tests positive, what is conventionally thought of as a false positive by the lay person. This number is negligible, something like 0.0001 IIRC.

    What we really need to know from PCR tests is when someone is infectious. I’ve read a number of articles about this and the consensus seems to be around below 26-28 cycles the detected virus can be cultured in to a live virus, anything above can’t. The last I saw was that PHE is doing 35-40 cycles of the PCR test analysis, so its likely that a number of people tested who were asymptomatic but had a test for some reason were still told to isolate if any virus was detected.

    As to false negatives, the LFD test again has a problem in that it only detects a high viral load, people who are around day 6 of their infection to day 12(?).

    There’s a good article on the pros and cons of those tests here:


  4. BND – thanks for that. I am not a viris sceptic or a lockdown sceptic, but I think it is important to know what the test statistics are actually telling us (on top of the use of tests in managing individuals in hospitals and T&T)
    The UK test positivity rate for the existing basket of tests is 0.4% right now. Even those who say that 0.4% means there is no Covid left should probably take note when it rises above that level, as it will.

  5. I’ve read elsewhere that as the prevalence of any disease in a population falls, the false positive rate becomes more of a problem when mass testing and ONS have also admitted that they were classifying a single gene indication in PCR tests when manufacturers requirements were for 2 or 3, so yet another reason for high UK reporting rates.

    I’ve also read that Dr Mike Yeadon and a lot of others advise that:
    “Healthy people under 60 are at a very low risk from the virus. If you didn’t seek a flu vaccine then you definitely don’t need this vaccine because flu risk is greater than covid risk for the under-70s.”

  6. @dio

    I think the PCR tests have a positivity rate of above 2% over the last week.

    We do not have a “small” number of cases at the moment, it only looks that way because we just had a big peak. The virus is still pretty prevalent – about 1 in 300 people according to the ONS random sample. That’s probably what you want to be looking at rather than the daily testing numbers.


    There have been times when the proportion testing positive was far far lower – in the ONS July surveys when the virus was under better control, it was more like one in 4000. Current numbers are over ten times higher than that! So absolutely no way this is a “casedemic” being driven by false positives from increased testing (and if increased testing bothers you just ignore the daily testing entirely and look at the ONS and REACT random sampling, but turns out that tells the same story). @BiND raises the more subtle point that PCR positives may not tell you exactly what you want to know, but the trend in PCR results should still give you a handle on the trend in the spread of the virus.

  7. According to the Zoe app, the cases in my area, and a lot of the rest of the country are 300/million – 0.03%.So that is not far from the 0.025% you report for last June. That means about 50 cases in my local authority. There are authorities with more cases but it does suggest that the virus is on the decline for now. So false positives are important

  8. @Dio

    Yeah, I’m not saying false positives are of no importance at all, and clearly some places have far lower prevalence than others. I’m saying it can’t be – at this still elevated level of national prevalence – a driver of national trends, which is what the “casedemic” folk talk up a lot. You’re switching between Zoe and ONS figures though which might not be the best comparison – they both give a picture of trends, but their estimates of levels aren’t actually terribly well calibrated with each other. In fact even ONS and REACT, which have pretty similarly methodology, agree far better on trends than levels. You were talking about PCR rates so it might be better to avoid looking at Zoe. If you look at the ONS data set, they do have a sub-national analysis (though with greater uncertainty than their national analysis for obvious reasons, and their sub-regions are pretty large) and there are a few areas at about the 1 in 1000 level, though not many. About 0.2% seems much more common even in the predominantly rural areas where prevalence is lower.


  9. @Dio

    In case you’re still about… you were having a brief discussion with @philip about a week back, regarding very long-term (historical scale) changes in interest rates. I plonked a reply in you might be interested by (has a link to the kind of paper I think you might enjoy reading) but was a bit late to the party so you might have missed it.

  10. I did indeed see that paper. It was intriguing but I was not fully persuaded. I wonder just how much you can rely on long term historical stats from periods where such figures did not exist in the same sort of way as they do today. The world economy, even national economies, are linked together today in ways that simply couldn’t exist before the middle of the 19thc. And, given that their research contradicted Piketty, the thought occurs that maybe researchers, whether consciously or not, somehow manage to find results that match what they were hoping to find

  11. @dio

    Yeah. Similarly, the thing I’m not persuaded about is “why should an economic trendline from 1100 to 1500 (or whatever) be relevant to predicting interest rates later in my lifetime?” Economic time series data is problematic for all kinds of reasons: very short timescales (modern GDP and growth data only goes back a few decades etc, everything is earlier is reconstructions of arguable accuracy), measures becoming targets then ceasing to work as measures (either because the figures get meddled with/redefined to look better, or because policies start targeting the measure rather than the underlying phenomenon it was supposed to be reflecting), even the fact that relationships between variables change over time so you have to stick to recent data to see the current state of play (at the expense of leaving you a puny sample size, making it harder to tell which apparent relationships are utterly spurious). It isn’t obvious to me how turfing up a bucketload of medieval data is going to resolve modern disputes about appropriate policy responses. Like you say the context and policy toolkit have changed so much.

    What I will say in favour of that paper, is that it has made me question any assumptions that interest rates had historically been stable or at least varied up and down within a steady range.

  12. It has made me wonder about the concept of a national interest rate before some very recent time. For example, if you were doing foreign trade in the 1850s, you would use bills of exchange drawn on a network of family banks or, if you lived in London, acceptable to the Acceptance Houses. The interest rate would have been specific to each transaction, based on the house’s perception of the risk of the deal. Similarly I imagine that a bank extending a mortgage would have based it on an assessment of the mortgagor’s risk profile, rather than on Libor plus 2,or some other formula. The story of Shylock comes to mind. Without doing any research, I wonder if our view of national interest rates is relevant before governments started funding wars with large scale debt issues. We had the Consolidated fund in the 19thc but coupons were fairly static for 100 years, I believe. Naybe the WW1 war loans were the real starting point

Leave a Reply

Your email address will not be published. Required fields are marked *