A few days back I commented somewhere here on DK that we would see 190K as a 7 day moving average for daily new COVID cases "early next week". I was thinking we’d hit that number after Monday’s data was mostly in.
Well, we're already here, after Saturday’s data — and 3 or 4 states didn’t report. I’m back to being a day or 2 behind the data in my thinking/calculations it seems. Quibble about if this is the Thanksgiving Surge, I think it's started. Someone posited that it would start early due to travel before Thanksgiving. Bad bet to take that is starts to slows down appreciably soon.
Pre-Thanksgiving, the 7 day moving average data fit a logistic curve extremely well. The average of 5 models compared to the real data shows the dip at Thanksgiving, and that the models and reality met at about 12/1 - 12/2.
I had hoped a surge would be delayed 3-5 more days and maybe the baseline would get to drop a little (and test my models a little more)...that would have been still a surge starting at 140-150K cases a day. As it is this current surge looks likely to have 165K as it's baseline.
This initial rise is probably enhanced from last week's dip, so we're probably a week away from me even trying a new spike analysis starting at 12/1/2020. Some preliminary grousing with the numbers is not encouraging. The last 3 days have averaged over 9K a day increase in the 7 day MA. Now, this is probably inflated, but I fear not by that much.
Eyeballing the worst part of the summer surge, I highlighted June 16 through July 17. Straight rise over run gives us an increase of about 1440 cases a day; the more careful linear regression estimates the slope a bit higher, so for numerical spit-balling 1500 looks like a decent estimate.
Looking at 10/19 through 11/18 (also 31 days), the calculated rise over run slope is about 3400 cases per day, and again the linear regression gives a slightly higher estimate. Let’s use 3500 as a compromise.
The above charts illustrate a bit of a problem with graphs — the different scales hide the fact that this recent surge was far worse than the summer surge since the slopes look similar at a glance. So let’s plot them together:
Now what we have seen before may not have any bearing on future events, but...we are looking at another 6 weeks of basically unchecked spread before a new Administration tries something…
If there is a relationship between the “baseline” of a surge and the slope of it’s growth, we’re in for a really bad time. The fall surge started at about 3 times higher than the summer surge, and had about 2.4 times (3500/1500) the slope. That would suggest that this current surge might be in the 8-9K a day range increase. The last few days have exceeded that (not surprising since the “replacement” level for the rolling 7 day moving averages this week are the depressed Thanksgiving week numbers), but if we have 10 days of 8K cases a day growth starting now, we’re butting up against 2 million cases a week. If the current “new surge” is about the same slope as the “first fall surge”, maybe we don’t quite make that scary milestone.
Stay safe. This looks like it is going to get much worse in the near term.