From the New York Times:
So I went there:
June 25: New Jersey began reporting probable deaths.
June 30: New York City released deaths from earlier periods but did not specify when they were from.
July 27: Texas began reporting deaths based on death certificates, causing a one-day increase.
...
In data for the United States, The Times is now including cases and deaths that have been identified by public health officials as probable coronavirus patients.
[But there is no explanation of when the Times began including probables.]
Some states and counties only report figures in which a coronavirus infection was confirmed through testing. Because confirmed cases are widely considered to be an undercount of the true toll, some state and local governments have started identifying probable cases and deaths using criteria that were developed by states and the federal government.
Confirmed cases and deaths are counts of individuals whose coronavirus infections were confirmed by a laboratory test. Probable cases and deaths count individuals who did not have a confirmed test but were evaluated using criteria developed by national and local governments. Some governments are reporting only confirmed cases, while others are reporting both confirmed and probable numbers. And there is also another set of governments that are reporting the two types of numbers combined without providing a way to separate the confirmed from the probable. The Times is now using the total of confirmed and probable counts when they are available individually or combined. Otherwise only the confirmed count will be shown.
Governments often revise data or report a large increase in cases on a single day without historical revisions, which can cause an irregular pattern in the daily reported figures. The Times is excluding these anomalies from seven-day averages when possible.
So...Confirmed+probable whenever possible. How about Hopkins? So, the 1,696 Deaths on July 27 are artifact of Texas' switch to death certificates. But the Times is not counting Texas' anomaly into its 7-day line, right? "When possible." I think it was not possible for the Times to exclude the Texas anomaly, I think the Texas anomaly is included in the Times' 7-day average because the line jumps too steeply, almost vertically. Now, there was a huge un-anomalous spike in reported Deaths on July 28, 1,324, but methinks it self-evident looking at the Times line after other big spikes that July 27's numbers are included. Look for instance at the line over May 4-6: a pimple on May 6 but otherwise the line average went down; manifestly the Times did find it "possible" to exclude New Jersey's June 25 spike, nary a blip is registered in the 7-day average; New York City's June 30 dump similarly is not included in the 7-day average, I think that is included. But Texas' anomaly is for certain included. You see nothing like the near-vertical rise in the average line after any previous spike,whether artifact of reporting or not.
From Johns Hopkins:
NYT's number has been higher than Hopkins for a couple, maybe three, days. Hitherto it had always been slightly lower.
From the Times:
So I went there:
June 25: New Jersey began reporting probable deaths.
June 30: New York City released deaths from earlier periods but did not specify when they were from.
July 27: Texas began reporting deaths based on death certificates, causing a one-day increase.
...
In data for the United States, The Times is now including cases and deaths that have been identified by public health officials as probable coronavirus patients.
[But there is no explanation of when the Times began including probables.]
Some states and counties only report figures in which a coronavirus infection was confirmed through testing. Because confirmed cases are widely considered to be an undercount of the true toll, some state and local governments have started identifying probable cases and deaths using criteria that were developed by states and the federal government.
Confirmed cases and deaths are counts of individuals whose coronavirus infections were confirmed by a laboratory test. Probable cases and deaths count individuals who did not have a confirmed test but were evaluated using criteria developed by national and local governments. Some governments are reporting only confirmed cases, while others are reporting both confirmed and probable numbers. And there is also another set of governments that are reporting the two types of numbers combined without providing a way to separate the confirmed from the probable. The Times is now using the total of confirmed and probable counts when they are available individually or combined. Otherwise only the confirmed count will be shown.
Governments often revise data or report a large increase in cases on a single day without historical revisions, which can cause an irregular pattern in the daily reported figures. The Times is excluding these anomalies from seven-day averages when possible.
So...Confirmed+probable whenever possible. How about Hopkins? So, the 1,696 Deaths on July 27 are artifact of Texas' switch to death certificates. But the Times is not counting Texas' anomaly into its 7-day line, right? "When possible." I think it was not possible for the Times to exclude the Texas anomaly, I think the Texas anomaly is included in the Times' 7-day average because the line jumps too steeply, almost vertically. Now, there was a huge un-anomalous spike in reported Deaths on July 28, 1,324, but methinks it self-evident looking at the Times line after other big spikes that July 27's numbers are included. Look for instance at the line over May 4-6: a pimple on May 6 but otherwise the line average went down; manifestly the Times did find it "possible" to exclude New Jersey's June 25 spike, nary a blip is registered in the 7-day average; New York City's June 30 dump similarly is not included in the 7-day average, I think that is included. But Texas' anomaly is for certain included. You see nothing like the near-vertical rise in the average line after any previous spike,whether artifact of reporting or not.
