CV and mad panic behaviour

[mateinone]

Hi mio.

Long time.  I largely agree with a lot of what you've written.  Can I ask why you have opted for 7 or 10 days vs 14?

Its hard to disagree with what you've written.  I think we ran out of the first lockdown too fast and paid for it.  We eased 3 times in 3 weeks with 7 days as the target, and it went backwards.

G'day Thry.
I come to the 7-10 day timeframe because it was the timeframe that allowed me most consistently predict trends in an accurate manner, whilst also keeping the data relevant.
I have put every countries daily data into a database, along with their population sizes etc and I ran models using  3-12 days trends, I found 7 was pretty much my sweet spot, of course the change even in the 7 day trend doesn't become apparent until around 14 or so days after a change to restrictions has been made and this is due to incubation periods, testing and reporting times etc.

14 days fails on the way up and the way down, on the way up, it does not come close to providing you with data in a timely manner (just as a small example we hit 50 cases over 7 days as an average a full 4 days before we hit that same average over 14 days). The problem is you are always making decisions long after you should have trended the data.
On the way down, it is the opposite effect. Knowing that we already know the data 10 days after a date (or approximately the data) off 7 days previous data, it makes no sense to "hang on" to data that is so far out of date.
For example the data across the previous 7 days is showing about a 20% drop compared to the last 14.

So on the way up and on the way down, it doesn't keep pace with the changes in the spread of the virus well enough for my liking and at times can really misrepresent the situation, but most dramatically, this is in the ascendancy

Obviously I am not an epidemiologist, but I just treated these stats the way I always treated stats and looked at this purely from a numbers perspective.