Overdue JFF model update (and why I've been totally screwed with incomptent data handeling).
Okay so it turns out the model isn't the problem after all, its the ridiculous way PHE has been putting out its statistics (and this is even before you get into all the arithmetic errors on cumulative data). Anyway essentially this is what has happened.
Before 27th March: Only 'most critical care workers' and patients were tested. This is described as 'pillar 1' in the daily brifings.
After 27th March: Testing capacity exists for both Pillar 1 and pillar 2 which includes other key workers.
So this might sound fairly straight forward. Up until the 27th march testing was saturated and I needed to weight by number tested. Whereas after it wasn't saturated so I could drop the number weighting and simply look at pillar 1. That would keep everything nice and consistent.
But oh no, it can't be that simple.
Because After 27th March BUT before 10th April: PHE was reporting Total i.e what you would intutiviely think as Pillar 1 + Pillar 2 as PILLAR 1 ONLY. I mean FFS. Notice how the model screwed up around the 10th April. This is why. Because either everything after 27th march but before 10th april was an underestimate (since it was not including Pillar 2 data) or everything after 10th April was an overestimate because it was including pillar 2 data.
So what happens when you try to fix this?
Well it looks a bit better.
Its not quite there though because I need to think about the saturation issues.
If testing is saturated then I weight by testing, if it isn't saturated then I don't weight. So this problem still needs sorting out.
I'll deal with this and hopefully have things back on track tommorow ish.
2023/2024 Snow days (approx 850hpa temp):
29/11 (-6), 30/11 (-6), 02/12 (-5), 03/12 (-5), 04/12 (-3), 16/01 (-3), 18/01 (-8), 08/02 (-5)
Total: 8 days with snow/sleet falling.
2022/2023 Snow days (approx 850hpa temp):
18/12 (-1), 06/03 (-6), 08/03 (-8), 09/03 (-6), 10/03 (-8), 11/03 (-5), 14/03 (-6)
Total: 7 days with snow/sleet falling.
2021/2022 Snow days (approx 850hpa temp):
26/11 (-5), 27/11 (-7), 28/11 (-6), 02/12 (-6), 06/01 (-5), 07/01 (-6), 06/02 (-5), 19/02 (-5), 24/02 (-7), 30/03 (-7), 31/03 (-8), 01/04 (-8)
Total: 12 days with snow/sleet falling.