Thread by @connolly_s

So Ireland has had 4 waves of PCR positives. The first three waves look wave shaped. The latest looks like it has been artificially constrained somehow... this could be testing capacity, biased sampling, etc. Need not be deliberate either... but doesn't look like the prior waves

If I look at the shapes of the prior waves, the current wave lost its shape around July 22nd... this is the point at which approx 50% of the population were fully stabbed, 58% had two stabs and needed to wait a week while 67% had had their first stab

Now the thing to note about Ireland is that the official government advice is that fully stabbed people do not need to get tested if they are a close contact of a confirmed case, unless they develop symptoms www2.hse.ie/conditions/cov…

So if more than half the population are excluded from testing, what would happen to a rising wave, especially if stabbing didn't prevent you from getting infected or infecting? Would that change the shape of the wave?

Another interesting anomaly is Sept 15th... and Sept 22nd seems very much like there was a delay in reporting some of the tests as the jump on the 22nd seems to match the dip on the 15th

Another interesting comparison is with this time last year. We had 686 cases on Oct 12th 2020 and this year we have 1475. What is the difference? Last year most things were shut down for everyone... this year it's only shut down for the non-stabbed... and they're blaming this on?

Now let's be fair, we are testing more people now than last year... 3.9 tests per 1,000 this year vs 2.7 tests per 1,000 last year, so let's (worst case) scale up last year's 686 * 3.9 / 2.7 = 990
This year is therefore at least 50% worse than last year despite all the stabbings

BUT the question now becomes what proportion of those being tested are stabbed, partially stabbed and non-stabbed? With 74.5% of the population fully stabbed, if we now direct most of the testing to the non-stabbed 25%, would that be a biased sampling?

Ok now in December I was looking at the initial data (before the test capacity anomaly on Dec 24th) and using a Gompertz fit of the first 6 days you get a very good model of how the wave panned out... keep in mind this was fitting just on the data up to Dec 23rd

We can do the same thing for July (this was an eyeball fit not an R^2 fit like the previous tweet, but again just the first 5 days rising up, I'm just reusing some earlier analysis routines to see what looks strange)
WARNING: this is a purely speculative fit, no stats support it

The interesting thing is that there looks to be about 2/3rd of this THEORETICAL fit's cases "missing". This is in contrast to, e.g. the residuals for March 2020 where we didn't have test capacity or the residuals for Dec 2020 which are cancelled out exactly by Jan 2021

So what if everyone is just as likely to get infected irrespective of whether you are fully/partially/non stabbed but testing is directed only at the non-stabbed? You'd expect a large number, between 50 and 75%, of the positives to be undetected? An interesting hypothesis?

Oh yeah... if that speculative fit is correct and all things are seasonal... then one would expect a 5th wave in late 2021/early 2022 that infects close to 1 million of the population irrespective of stab status... let's hope my speculative fit based on 5 days in July is wrong!

But remember my speculative fit based on the 5 days before Dec 24th and the Gompertz rate constant from the Oct 2020 wave was accurate to within 1% for the Dec 2020-Mar 2021 wave... the first 5 days can be scarily accurate!!!

Oh and before I forget, the December fit was from before lockdown and by the 26th we were in the strictest lockdown ever… and the wave still followed Gompertz… so if you think masks and lockdown did anything one way or the other, you might want to rethink with an open mind!

@threadreaderapp unroll