So, the latest weekly update by the CDC came in at 1168 cases, 80 more than the previous one.
Admittedly, this weekly number of 80 seems way too high, and the numbers don't seem to add up:
Today (June 6), Texas reported that "Cases have not increased since the June 3 update", and that "As outbreak-related cases continue to slow, DSHS will provide measles updates on a weekly basis each Tuesday, beginning June 10" (until today, updates were posted twice per week, on Tuesdays & Fridays), i.e. confirming that the outbreak there is at its end tail
The situation in New Mexico and Oklahoma seems largely stable, with weekly new cases in the (low) single digits.
Kansas went from 64 cases on May 28 to 71 cases on Jun 4, i.e. just 7 new ones
North Dakota went from 28 cases on May 29 to 34 cases on June 6, i.e. again just 6 new ones
The number of jurisdictions reporting cases to CDC went from 33 to 34; the new jurisdiction is South Dakota, which however seems to have reported one single case on June 2 (no detailed data in the state DoH site)
So, where the heck did all these 80 new cases come from?
The answer has been hinted at below by @allthingsthatare and @doodlebug: the vast majority of the 80 "new" cases actually come from revisions of past data, i.e. delayed reporting, which actually goes as far back as early March!
We can see the previous (May 30) version of the CDC page archived at the Wayback Machine; here are the revisions to their weekly data (numbers reported prior to Mar 9 seem to be unchanged):
Week
Reading on May 30
Reading on June 6
Difference
Mar 9
83
85
2
Mar 16
102
104
2
Mar 23
89
90
1
Mar 30
116
116
0
Apr 6
74
75
1
Apr 13
83
86
3
Apr 20
43
49
6
Apr 27
44
49
5
May 4
30
34
4
May 11
15
24
9
May 18
17
45
28
May 25
20
37
17
June 1
n/a
2
2
So, it is apparent that:
The vast majority (78) of the "new" cases (80) are actually due to delayed reporting of old ones, from as far back as early March
More than half of the 80 "new" cases (45) are due to two bulk updates of May 18 & 25
A good 1/4 of the "new" cases (20 out of 80) actually come from March & April updates
What's more, save for (arguably not unexpected) updates of 2-3 weeks prior, such updates (bulk and/or going that far back) seem not to be present in older snapshots of the page (see e.g. the May 23 snapshot), although we can find past weeks which were later updated lower by 1 case.
In forecasting lingo, we call such situations data bombs, and they can make the life of us forecasters significantly more difficult...
Files
Why might you be wrong?
As before, plus the additional dangers coming with such data bombs...
Thank you for this! While updates all the way to March-April are unusual (and, frankly, inexcusable, whoever fault it is), the updates to weeks -1 to -3 are rather normal, if my memory from 2024 serves. So in any case we seem to have the kind of +40 uptick, which is troubling for the idea that the current outbreak is nearly over. I'll wait another week to see how this analysis holds before I update. Going by 2024 rate, in the remaining months we can expect +166. To get to the lowest bin is +188. Given the current rise, even if it fizzles relatively soon, that most likely means that the lowest bin is out of the consideration.
Why do you think you're right?
So, the latest weekly update by the CDC came in at 1168 cases, 80 more than the previous one.
Admittedly, this weekly number of 80 seems way too high, and the numbers don't seem to add up:
So, where the heck did all these 80 new cases come from?
The answer has been hinted at below by @allthingsthatare and @doodlebug: the vast majority of the 80 "new" cases actually come from revisions of past data, i.e. delayed reporting, which actually goes as far back as early March!
We can see the previous (May 30) version of the CDC page archived at the Wayback Machine; here are the revisions to their weekly data (numbers reported prior to Mar 9 seem to be unchanged):
So, it is apparent that:
What's more, save for (arguably not unexpected) updates of 2-3 weeks prior, such updates (bulk and/or going that far back) seem not to be present in older snapshots of the page (see e.g. the May 23 snapshot), although we can find past weeks which were later updated lower by 1 case.
In forecasting lingo, we call such situations data bombs, and they can make the life of us forecasters significantly more difficult...
Why might you be wrong?
As before, plus the additional dangers coming with such data bombs...