The positivity trend in Scotland has a distinct correlation with weekends.
During weekdays test numbers are high but at weekends they drop by 4 or 5 thousand, and the positivity increases by several percent.
Good work, but it's banal and should be obvious.
The positivity increases at the weekend because ttests done at the weekend contain a higher proportion of people who need tests clinically, fewer optional tests, e.g workers etc. If a person is sick he needs testing, if not sick he does not , hence if a person is sick he'll be positive more often.
It is sad that essential clinical need tests are mixed in with routine tests.That is a banal data issue. It makes it impossible to spot real trends. The medics have shown themselves to be unable to read data. They might be stupid.
There is this message from PHE. you find it in this wiki:
https://en.wikipedia.org/wiki/COVID-19_pandemic_in_the_United_Kingdom
On 3 October, Public Health England announced that a 'technical error' had caused under-reporting of new cases for recent dates, and that the missing positive results would be declared over the forthcoming days. The number of new cases declared on 3 October was approximately double the rate prevailing over the preceding few days.
If you find a set of -ve and +ve results that you left out of your data, and 'fixed it' by only declaring the missing positive results as PHE said they would, the effect would be to alter the ratio of -ve to +ve recotds in the data, so around Oct 3, you would have a relative surplus of +ve results, giving the impression of increased positivity, as you snapshot shows. It looks like the data they had was not timestamped, so excess records got clumped on Oct 3.
It is deeply shameful if data has been so abused by PHE and others. They have no idea how to do things properly. If you have a data set, and some section has gone missing, it can only be repaired by accounting for all the records, not just the +ve ones. PHE is staffed by cretins so it appears.
My advice, drop records for a few days either side of Oct 3, they cannot be trust for data trends.
I've finished the scripting now. Sample date as somebody suggested didn't produce a data set when I tried it but I can revisit that. I have written a php script that pulls down the data as a CSV then produces a chart. I've applied 7 day smoothing and log scale as defaults but they can be switched on and off at will. The glitch at the star of October is very apparent and the higher % +ve continues thereafter.
This just runs on a Rasperry Pi on my home broadband so please don't post the link anywhere else or I will have to password protect it.
Does anyone know what happened to the BL2 Online Records Spreadsheet? Many players and records were removed and all of the links to record replays are gone.
There is this message from PHE. you find it in this wiki:
https://en.wikipedia.org/wiki/COVID-19_pandemic_in_the_United_Kingdom
On 3 October, Public Health England announced that a 'technical error' had caused under-reporting of new cases for recent dates, and that the missing positive results would be declared over the forthcoming days. The number of new cases declared on 3 October was approximately double the rate prevailing over the preceding few days.
If you find a set of -ve and +ve results that you left out of your data, and 'fixed it' by only declaring the missing positive results as PHE said they would, the effect would be to alter the ratio of -ve to +ve recotds in the data, so around Oct 3, you would have a relative surplus of +ve results, giving the impression of increased positivity, as you snapshot shows. It looks like the data they had was not timestamped, so excess records got clumped on Oct 3.
It is deeply shameful if data has been so abused by PHE and others. They have no idea how to do things properly. If you have a data set, and some section has gone missing, it can only be repaired by accounting for all the records, not just the +ve ones. PHE is staffed by cretins so it appears.
My advice, drop records for a few days either side of Oct 3, they cannot be trust for data trends.
I did a little more work and put the percentage figures into a separate graph so both log and linear can be used as the huge difference in scale between the numbers is removed and in the linear plot it seems obvious to me that your explanation is the correct one. Thanks for clearing it up. I had suspected something more sinister than simple incompetence. Perhaps they need the good Professor Ferguson to advise them on good coding and IT practice. Or maybe not 🙂
Forbidden
You don't have permission to access this resource.
Apache/2.4.38 (Raspbian) Server at covid.digitalham.co.uk Port 443
