Arg. more distress and errors. Contrary to: https://www.washington.edu/news/2020/04/10/covid-19-peak-active-cases/ which says first differences peaked on April 5-7. I also recalculated the fits from yesterday, and it looks like I was being overly pessimistic about US situation (and may have had an index error). Here is the revision: https://0x0.st/iSFY.png In any case, estimated bin-widths probably have an error +/-2 at 95% confidence. We still don't know how long the tail will be, it's just a guess, and a work in progress. --Brad On Sun, Apr 12, 2020 at 11:47 AM Brad Klee <bradklee@gmail.com> wrote:
Using data up to today, total_cases.csv from: https://ourworldindata.org/coronavirus-source-data
U.S. Data: ... 6427, 9415, 14250, 19624, 26747, 35206, 46442, 55231, 69194, 85991, 104686, 124665, 143025, 164620, 189618, 216721, 245540, 277965, 312237, 337635, 368196, 398809, 432132, 466033, 501560, 529951
Second Differences: {1847, 539, 1749, 1336, 2777, -2447, 5174, 2834, 1898, 1284, -1619, 3235, 3403, 2105, 1716, 3606, 1847, -8874, 5163, 52, 2710, 578, 1626, -7136}
There is a systematic pattern of negatives corresponding to week ends, which suggests a bin-and-average:
Mean[{ 5174, 2834, 1898, 1284, -1619}] = 1914 Mean[{3235, 3403, 2105, 1716, 3606, 1847, -8874}] = 1005 Mean[{5163, 52, 2710, 578, 1626, -7136}] = 499
Or
Mean[{-2447, 5174, 2834, 1898, 1284}] = 1749 Mean[{-1619, 3235, 3403, 2105, 1716, 3606, 1847}] = 2042 Mean[{-8874,5163, 52, 2710, 578, 1626}] = 209
Either way it looks like US is not yet on peak of daily new cases, contrary to some reports, ex: