Thread by @plaforscience

RELEVANT RESULT

We've found that False (Old) PCR positivity is a function of PCR average Ct.
Fake Old Cases are created if average Ct is over ~27.5
Then on, Fake proportion grows at ~30% for each average Ct point.
At AvCt>31 we will have ~only old fake cases.

The method:
THREAD

We have Madrid dataset with PCR&Antigen test results and PCR Ct averages.

Using them we can calculate the positivities found with each of the methods. We observe they're similar, but PCR is higher with lower prevalence, and Ant with higher prevalence.

Let's compare them.

If we find their proportions we notice PCR is almost all the way giving MUCH HIGHER positivities than Antigen.
They could have same rate (1) if test were technically equal, OR a constant disproportion due to better specificity/sensitivity.
But changing over time makes NO SENSE.

As we deeply know the problem, we were familiar with the resulting unexplainable curve. You've seen it in the first tweet: the Average Ct (blue)

If we put both curves together, you CAN see the relation.
It does makes all sense that extra PCR positivity comes from a test issue.

It means Fake Positivity is a function of Ct.

We calculate it by considering there's a TRUE PREVALENCE, the one u'll find if u had a PERFECT measure.
We defined equations considering all factors involved:
True positives, false positives, old positives and false negatives.

As TRUE positivity is common, nomatter test type, we can solve the equations and find it, and the burden of extra offices PCR is showing.

A trøll explain will be different sensitivity/specificity is to blame.
We checked it out under same equations.

Spoiler, trølls are wrong.

There're ranges for PCR& Antigen sensitivities and specificities.
We tried all limit and average combinations

It didn't change.
PCR detection NEVER was proportional, there was always UNEXPLAINABLE VARYING EXTRA PCR CASES.
And they were ALWAYS f(Ct)

Old viral fragments detection

There were different local anomalies (quit deep at extreme assumptions) depending on values for specificity and sensitivity, but they NEVER lost they Ct trend

We set sensitivity& specificity most accepted values and calculated the function relating Average Ct value and Old Cases

We isolated the needed fraction of PCR extra positivity that made true antigen&PCR positivities equal, and related it with average Ct value.

With it you can calculate true PCR+, OLD PCR+ and true Antigen+; and ask associated data: positivities, number of true cases...

Going back to this graph, in our model proportion between PCR& ANTIGEN TRUE positivity (& I mean TRUE, the exact ones almighty Newton knows) is perfectly flat at 1.
Official measure is bumping like crazy with (you've seen it's not sensitivity/specificity) no rational explanation

With median equation solutions (K) from data, we modelled an approximate general equation:

%Fpos=(AvCt - 27.5)/3,42

%Fpos: percentage of old fake PCR+
AvCt: Average Ct value.

It's only an approximate, but it fits well in True Positivity and observed old positivity.

Of course Average Ct is lacking a lot of scale, and median approximation, finesse; but if we check PCR/ANTIGEN positivities, we check out K model is much closer to fit in flat ~1

Max deviation is ±50%, with average ~0
Official is up to +150% with average over 50%!!

We've plan to use the disaggregated by Ct ranges for s better refinement.

We wanted a first, even more imprecise, model relaying only in Average Ct, as we know Ct data is so damn difficult to find. The more general the data needed, the more chance we could use it somewhere else.

Just checking with the model we already made from true Epidemic from Ct ranges (<20,21-29,>30) we notice trends were correct, and more refine can be done.
Just when we find time.
Patience.

The govt has ALL the data.
Why don't they simply DISCARD fraudulent old PCR?!

We think this is so RELEVANT, and we're asking for anyone to bring any data series which include Ct values, si we can accurate the model

By now, this is mathematical proof for the existence of HIGH proportion PCR Old fake positivity, SPECIALLY in low prevalence environments.

🦆