The kind of nonsense below isn't born out by the data. Since I'm about a week from getting my Masters in Data Analytics I have been running a model on the outbreak as part of my capstone. It turned out a little different than others you'll see. I'm not forecasting deaths. I've just built a dashboard about current state that gives some insight into what is happening now based on the last few weeks. You can use it to take educated guesses on how mitigation has worked. It was inspired by how I track infections in Cybersecurity Incident Response to enable decision making. As a caveat, my numbers aren't 100% the same as others you'll find because I've pulled them from multiple sources. They will be similar though.
So what is the reason I am rolling my own? Many of the sites had different reporting gaps and collated their data differently and in sometimes bad ways imo. Sometimes they are being forensic and not backdating data, other times it looks like laziness. For example, covidtracking.com had missing data for awhile when not all states were reporting.
newsbreak.com actually had solid data but they like many others just dumped NYs revised numbers in on one day. That's fine if you want to just keep a simple tally but it breaks your growth curves. I personally went back to NYs revised data which was by day and fixed mine to reflect the re-statement. Anyway, I won't say its perfect but it is good enough.
According to my numbers (which closely track the best to the newsbreak numbers), new cases didn't peak on the 4th. Instead, the highest count of new cases so far was the 10th of April. There was a slight dip for a few days but that corresponds with the time that labs were saying that they had idle capacity (sometimes 30% capacity unused by some accounts). This has been ascribed by some as tight CDC guidance about who to test which artificially lowered numbers across the board. In any case, the number of new cases on the 16th was within a few hundred of the 10th. Aka a double top. The number of new cases has been moving horizontally for over a week. That has made the count of total cases from the period 4/7 - 4/16 essentially linear. Linear isn't how nature works. So some limit is there. It could be social distancing but that probably isn't going to keep a perfect linear lid on it. I'd suspect hitting a capacity cap on number of tests completed. It's an easy explanation considering what we know.
Here are the 14-day moving windows for new cases:
* period 2/24 - 3/8 - growth of new cases was 18% / day - doubling every 4 days
* period 3/9 - 3/22 - growth of new cases was 29% / day - doubling every 2.5 days
* period 3/16 - 3/29 - growth of new cases around 26% / day - doubling about every 3 days
* period 4/4 - 4/17 - near linear growing at 6% / day - doubling every 12 days <- social distancing? testing cap? suspiciously low imo.
As to the idea that the new death's count is lagging that is true but the rate of new deaths is still growing. If you take 14-day moving windows:
* period 3/9 - 3/22 - growth of total deaths was 22% / day - doubling every 3-4 days.
* period 3/16 - 3/29 - growth accelerated to 26% / day - doubling about every 3 days.
* period 4/4 - 4/17 - near linear growing at 10% / day - doubling every 7 days - supports social distancing
That means recent growth in total deaths is much slower than early in the outbreak but obviously still horrible. It means social distancing worked most likely.
Some conclusions I can make. Cases are likely increasing and testing is missing them. I don't think that is going to prove wrong. The number of new cases moving horizontally doesn't make much sense. It could be real but low confidence it is real. That means the number of deaths will continue to grow. When cases truly peak we'll know about 2 to 3 weeks afterwards. That implies we could have peaked but we have not seen that drop in deaths yet. I personally think it is unlikely we actually got to a peak in new cases. They are just declaring victory to keep the dear leader happy. So sorry for as usual stealing hope but that is only because it is probably the reality of it. Fuck. This is the worst timeline.
https://twitter.com/realDonaldTrump/sta ... 7696101376