I've been pedantic about many ZOE Covid Study criticisms. Perhaps unfairly, given ZOE's lack of clarity. This post is well-written as I'd expect, and hits really important limitations.
Both "accurate and ahead of other methods" is my personal bugbear. It was their repeated claim, while its use of a 14 day average was often not mentioned.
They took an average of swabs from the 14 days ending on Nov 10, and used that to judge a lockdown beginning on Nov 5?
They later wrote a Spectator article about it, which included the same judgement. I had a direct response from Spector himself. I still can't make sense of his reasoning.
That is all pretty odd yes (and sadly consistent with what I've seen from them elsewhere). Of course I'm sure all the statistics behind the nutrition stuff is much more rigorous.
Do you know if the ‘as it appeared in real time’ Zoe data are available anywhere? I remember there being quite a few inconsistencies (some of which you’ve helpfully documented here) so could be useful to have the option to compare performance systematically...
I'm not sure that they are - Alan (also in the replies) might know better because I know he kept a close eye on things, so maybe he (or someone else) archived stuff. There's a partial record in blog entries here https://covid-webflow.zoe.com/blog I think, but not in an easy to access format.
I have some older "incidence table"'s on my computer. The filenames I used were not quite as useful. All the England data from the old ones is in a spreadsheet "incidence table - England time series", which can be found here: https://drive.google.com/drive/folders/1O9vUWKufI3KfaGtNvKjBTKxHq6yNcyfX
If anyone wants more, please do ask. Caveat is, the incomplete archive on my computer is 16GB. I'm not planning to host everything publicly or indefinitely just for the sake of it.
I've been pedantic about many ZOE Covid Study criticisms. Perhaps unfairly, given ZOE's lack of clarity. This post is well-written as I'd expect, and hits really important limitations.
Both "accurate and ahead of other methods" is my personal bugbear. It was their repeated claim, while its use of a 14 day average was often not mentioned.
They took an average of swabs from the 14 days ending on Nov 10, and used that to judge a lockdown beginning on Nov 5?
https://web.archive.org/web/20201114105840/https://twitter.com/sourcejedi/status/1327566192253526017
They later wrote a Spectator article about it, which included the same judgement. I had a direct response from Spector himself. I still can't make sense of his reasoning.
https://gist.github.com/sourcejedi/0cf3df39c94ece6d379ba9e4d4e3eadd
That is all pretty odd yes (and sadly consistent with what I've seen from them elsewhere). Of course I'm sure all the statistics behind the nutrition stuff is much more rigorous.
Do you know if the ‘as it appeared in real time’ Zoe data are available anywhere? I remember there being quite a few inconsistencies (some of which you’ve helpfully documented here) so could be useful to have the option to compare performance systematically...
I'm not sure that they are - Alan (also in the replies) might know better because I know he kept a close eye on things, so maybe he (or someone else) archived stuff. There's a partial record in blog entries here https://covid-webflow.zoe.com/blog I think, but not in an easy to access format.
ZOE's "covid-public-data" files are documented here: https://sourcejedi.github.io/2022/01/31/zoe-covid-study.html . But you have to check the timestamps, because a few files were overwritten by retrospective corrections.
There's my download-sample directory. (Not complete). "incidence table" was a small file they never mentioned, hence easy to archive here:
https://github.com/sourcejedi/nova-covid/tree/main/download-sample
I have some older "incidence table"'s on my computer. The filenames I used were not quite as useful. All the England data from the old ones is in a spreadsheet "incidence table - England time series", which can be found here: https://drive.google.com/drive/folders/1O9vUWKufI3KfaGtNvKjBTKxHq6yNcyfX
If anyone wants more, please do ask. Caveat is, the incomplete archive on my computer is 16GB. I'm not planning to host everything publicly or indefinitely just for the sake of it.