Jump to content


Photo

Face Coverings?


  • Please log in to reply
442 replies to this topic

#436 mel2

mel2

    Virtuoso

  • Members
  • PipPipPipPipPip
  • 5351 posts
  • Member: 6928
    Joined: 15-May 06
  • East Yorkshire

Posted 21 September 2020 - 23:44

It was disappointing that the CMO and CSO did not take questions after their presentation. I would like to have seen the projected "not a prediction" graph challenged; I see no reason why the UK rates should not plateau as appears to have happened in Spain. It's hard to glean detail other than mere numbers from the gov.uk site regarding hospital admissions, i.e whether Covid is the reason for the admission or the patient tested positive on entry. A 5-week view of 'covid' hospital admissions reads thus: 49; 44; 67; 135; 205. An increase for sure, but hardly a 'doubling every week'. One would almost think we were being prepped for some tough extra restrictions to be announced tomorrow. I can't wait!
(I did read of someone with a gastrointestinal problem being placed in a Covid ward because there was plenty of room there, and being discharged 4 days later; not sure that it should have happened, or if it is likely to have influenced the figures)
I'd like to thank those who have taken the trouble to make sense of the FP issue - it looks like one of those rabbit holes that doesn't lead to a clear answer; all I know is it seems a very bad idea to have a test, for all sorts of reasons.
  • 0

#437 Hildegard

Hildegard

    Prodigy

  • Members
  • PipPipPipPip
  • 1444 posts
  • Member: 887389
    Joined: 26-October 13

Posted Yesterday, 07:18

It was disappointing that the CMO and CSO did not take questions after their presentation.

 

Probably inevitable, though, after press questioning tried to trap both officers into giving political views earlier this year.

 

I agree that UK infection rates could plateau as they have in Spain - although I doubt that anyone would welcome them plateau-ing on 10,000 new cases a day, as they have in Spain. It depends on how effectively UK citizens stick to the new rules, but I doubt that the UK will be willing to bring the army onto the streets to enforce lockdowns, as has happened in Madrid.


  • 3

#438 EllieD

EllieD

    Advanced Member

  • Members
  • PipPipPip
  • 903 posts
  • Member: 897806
    Joined: 04-June 17

Posted Yesterday, 08:13

 

 

b is the false positives

c is the false negatives

 

Usually the 'false positive rate' is definied as 1 - specificity.  So say you have 1000 people who have the don't have the disease, and 992 test negative with 8 testing positive.  You have eight false positives.  You therefore have

Specificity = 8 / (992 + 8) = 0.992 or 99.2%

Your false positives are therefore 100-99.2 = 0.8% (which you can also get by 8/1000; or you can call 0.008)

 

Thank you for taking the time for this detail Maizie! Me being thick, I'm sure, but just wondering, please could you say what numbers are being used for a, b, c and d? Your example says 8 test positive, but also that there are 8 false positives. if b=8, what is the value of a? Or are they both 8 in this example?


  • 0

#439 Piano Meg

Piano Meg

    Advanced Member

  • Members
  • PipPipPip
  • 520 posts
  • Member: 271350
    Joined: 14-June 11
  • Commuter country!

Posted Yesterday, 08:58

 So say you have 1000 people who have the don't have the disease, and 992 test negative with 8 testing positive.  You have eight false positives.  You therefore have

Specificity = 8 / (992 + 8) = 0.992 or 99.2%

Your false positives are therefore 100-99.2 = 0.8% (which you can also get by 8/1000; or you can call 0.008)

                      Reality
              Person has the disease    Person does not have the disease

Test result

Positive               a                           b

Negative               c                           d

False positives muck up your numbers; false negatives go around spreading disease :D

 

Thank you for taking the time for this detail Maizie! Me being thick, I'm sure, but just wondering, please could you say what numbers are being used for a, b, c and d? Your example says 8 test positive, but also that there are 8 false positives. if b=8, what is the value of a? Or are they both 8 in this example?

In this demo, the scenario is that no-one actually has the disease, so I think a=0. 

But best get confirmation of that from Maizie when she next logs on!

I'll second my thanks to Maizie too. Very helpful.

 

edit: so c must also be zero, and d would be 992


  • 1

#440 Maizie

Maizie

    Virtuoso

  • Members
  • PipPipPipPipPip
  • 6905 posts
  • Member: 9360
    Joined: 05-February 07
  • Bishop's Stortford, Hertfordshire

Posted Yesterday, 09:16

Spot on, Piano Meg!  Let me draw it out with the numbers from my example

                      "Reality"
             Has disease    Does not have disease
Test result
Positive        a                b = 8

Negative        c                d = 992

We were only looking at false positives, i.e. people who do not have the disease but test positive.  For this, you only need to know b and d.

a and c are essentially unrelated to b and d, and could be any numbers you care to make up.  This is why it's useful to know both the false positive rate (from b and d) and the false negative rate (from a and c) so you can get a full picture of a test's usefulness.

If you have all four numbers, you can also calculate the positive predictive power and the negative predictive power, which in English are "if I get a positive test result, what is the probabiIity I really have the disease?" and "if I get a negative test result, what is the probability I really don't have the disease?".  I think these are probably easier for people to understand and relate to but they're not necessarily the numbers we get told :rolleyes:

The enormous difficultly in this is that we need a test to tell us who has the disease or not but we also need to know what I've labelled "Reality" (i.e. who truly has the disease or not) to know how good the test is.  If you only have one test method, then you have your positive and negative results, but no idea how they relate to reality :blink:

This type of table is really useful if you are comparing a new test to an existing "gold standard" test (which you can consider to be "reality").  Or it's useful if you can compare "all the people who took a test at a certain time" to "retrospectively reviewing all those people months later to determine who was/wasn't ill at that time and deeming this reality".  With new tests for a new disease, there will be a lot of unknowns and a lot of frequently changing numbers as things become clearer (and hopefully settle down)


  • 0

#441 corenfa

corenfa

    Virtuoso

  • Members
  • PipPipPipPipPip
  • 6675 posts
  • Member: 95861
    Joined: 28-March 10
  • Here

Posted Yesterday, 12:27

More about "Long Covid"

 

https://www.bbc.co.u...tories-54106272


  • 4

#442 ma non troppo

ma non troppo

    Virtuoso

  • Members
  • PipPipPipPipPip
  • 2063 posts
  • Member: 76027
    Joined: 23-September 09

Posted Yesterday, 21:46

Another very interesting BBC commentary.

https://www.bbc.co.u...health-54255635
  • 0

#443 mel2

mel2

    Virtuoso

  • Members
  • PipPipPipPipPip
  • 5351 posts
  • Member: 6928
    Joined: 15-May 06
  • East Yorkshire

Posted Yesterday, 23:01

It was disappointing that the CMO and CSO did not take questions after their presentation.

 
Probably inevitable, though, after press questioning tried to trap both officers into giving political views earlier this year.
 

Not really good enough.
If they use naked scare-tactics like this with worst case scenarios they should have the courage of their convictions and be prepared to defend their reasoning.
Glad that throughout today many have seen it for what it is.
You were probably being less than serious about bringing the army in to enforce the decree - sorry, rules, but it looks like it may come to pass after all. Quite disproportionate. I hope Sir Graham Brady can exert some influence in the right direction.
  • 0