Thread by @RP131

I've been trying to track down the precise source for the "126.8" figure used in the document to recommend the Wales fire-break. I think it's important since they've taken the rare step of publishing the thresholds it was based on, so we can actually monitor it going forwards. 1/

It doesn't explicitly say here what this means other than "cases per 100k". My guess would be weekly cases since it sounds roughly right and that's what they often use. But I couldn't find an exact match in the Wales nation level data. Closest I could find highlighted below:

They do use the word "estimated". Could this mean they've calculated it somehow? Nothing in the supporting documents about this and just says they'll use "Number of confirmed COVID-19 cases" for the transmission circuit-breaker.

It also says the figure breached "40" since 17-Sep, which according to the 19-Oct data this measure did, so that adds confidence we're looking at the correct column.

In a section above it actually describes it as "rolling 7 day average":

But I don't see how they could have reached the number of "126.8" by using a 7 day average. To still be in the hundreds it would need to be an average of the 7 day totals and because these numbers are quite lagged, the averages all come out higher:

So my guess is the "average" bit is a typo, possibly due to them using the term "rolling" to mean a moving window but accidentally putting in the word "average" because it so often goes with "rolling"?

I also considered if it could be "reported date" data - but again I can't see how that would fit. Plus these numbers didn't breech 40 until about 21-Sep.

That section also mentions the period of 11th to 17th October, so that adds weight to my guess above that they meant the number that my figures show as "122.4". Perhaps they measure it slightly differently? Or maybe it's another typo?

I realise I'm being quite pedantic here and the purpose of this thread is not "haha they got the number slightly wrong" - it's a genuine attempt to fully understand the source of the information so we can be confident we're tracking the same thing.

We also need to be careful because if they did use that 2-day figure then it's subject to quite a lot of lag. As you can see reading further up the table, the settled numbers are quite a lot higher. We need to make sure we're comparing like-for-like going forwards.

There's another section in the document where they also quote rates per 100K numbers but this time they do match the data for the period they specify. So it's correct by region, then why not at nation level?

Congratulations if you've made it this far! I realise this is one of my more pithy analyses and not particularly exciting. Now's the point when someone just pastes a screenshot from the Public Health Wales website showing the 126.8 and I find I've wasted the last hour on this! 🥴