Perhaps someone can clarify something for me. In general, we are told that false positives are between 0.8% and 4.0%, median 2.3%. We don't actually know what they are at the moment.
So let's look at the cases/tests from the ONS figures (all 7-day averages to yesterday).
They were very flat at about 0.5% for all of July and August. Could we say that this is likely some sort of 'baseline' false positive rate? Is, indeed, the false positive rate likely to change much?
Now have a look over the last week or so. It is early to say, but it seems to be flattening at about 1.4%. In the first week of September it seemed to gradually ramp up from 0.5% to 1.4%. A whiff of rodents, I thought.
Over the last couple of weeks, there has been a stronger emphasis on these test results to drive the government restrictions. Was something, like a sensitivity, turned up in the first week of September?
If one follows the real meaning of false positives, we know that these test results are as useful as a turd in a punchbowl. In addition, if it is the case that a flattening of this curve represents the real false positive, any sort of manipulation of sensitivity can be sussed and ruled out.
I tried to look further at this an puzzled. Probably me, but:
: What is the calculation to get "thus a maximum of .57% of these cases..." ? Not obvious to me.
: The blue line looks to be cumulative. But how is it generated? Each day take a 7-day average then cumulate from the previous day's 7-day average?
I don't see enough information to discern how many "real" cases, e.g. accounting for false positives.
For the 0.57%, I just took the minimum of false positives (0.8%) from the case percentage (1.37%). That means there could never be any more real positives that 0.57%.
All the figures are 7-day averages. Take the average of 3 days before plus today plus 3 days after. That is why the chart doesn't go beyond three days ago.
My question is whether, when the curve of Cases/Tests is flat - regardless of all sorts of external variables - has it bottomed-out at a regular level of false positives? If so, that would have been about 0.5%, which is somewhat more accurate than the 0.8% that we were told. If so, then we would have to take 0.5% from 1.37%.
However, it appears that the recent measurements have flattened out (still early days to be sure) at 1.4%. So we could take 1.4% from 1.37% and, oops, we have -0.03%! It would be tantamount saying that there are no real positives out there!
Now I am not proposing that is really the case at this stage, but we need to keep an eye on this because it rubbishes the test program entirely.
Thanks. I see what you've done.
I still can't get my head around all this. Too much data, but not enough useful data. All down to not knowing the real False Positive Rates and False Negative Rates which I know will change over time. And perhaps more significantly depends on quality/spec of the test. All these are mysteries.
There are so many question I have and I've not found answers. The first one is of those "cases" how many are actually real? What is a case? Using UK Gov't nomenclature, what doe "lab confirmed" mean (i.e. explain how they confirm and is it done with multiple tests?)?
I've run across a website from "Health Data Research UK" https://www.hdruk.ac.uk/projects/false-positives/ which addresses the key issue. They even provide a spreadsheet tool to do the computations of probabilities of "truly" positive after one, two, and three tests.
Before continuing this conversation here ... I'm going to dig into this and perhaps you may wish to have a peak.
Undated document, but https://www.ouh.nhs.uk/working-for-us/staff/documents/staff-testing-privacy-statement.pdf from Oxford University Hospitals says
"The test is a swab of your nose and throat undertaken at one of the Trust testing sites. The test is not 100% accurate; at present it is thought to detect approximately 70% of coronavirus infections."
Uses the following for reference:
Comparative accuracy of oropharyngeal and nasopharyngeal swabs for diagnosis of COVID-19,
www.cebm.net/covid-19/comparative-accuracy-of-oropharyngeal-and-nasopharyngeal-swabs-for-diagnosis- of-covid-19 2020
Same time in recent Spectator column they say less than 1%. https://www.spectator.co.uk/article/covid-19-and-the-false-positive-trap
