The daily deaths numbers include an entry of deaths which were NOT on that day (notice the blue spike on the right side of the graph).
This ruins the trend line - which we desperately need, in order to make good decisions.
That spike is 1831 "re-classified" deaths from earlier by New Jersey, which were likely COVID-19, but not counted.
I took those 1831, and distributed them across the entire period of New Jersey's deaths. 2.09% of NJ's total deaths were reported on April 8, so I assigned 2.09%*1831 to that day as additional deaths.
The green line represents the adjusted account.
Since 1831 deaths is a pretty small slice of the nation's 121,000+, so the green line mostly overlies the blue one.
Then I ran the 7Day average on those adjusted numbers - notice that the dark red line also mostly overlies the lighter red line of unadjusted numbers.
This removes the spike, restores the trend, provides the correct number (sort of) of actual deaths, and makes at least a rational guess as to when those deaths happened.
The only REAL way to fix it is to retrieve the day for every single one of those 1831 deaths, one by one, and re-assign them to their actual days.
Short of that, this analysis is better than having a misleading spike.
There is also ANOTHER similar spike from June 30, but this dataset did not record that one. Good news: There is still a SLIGHT decline visible in the adjusted 7-day trend.
Bad news: This is our 'grace period' in the national data, before the deaths, which lag behind cases by about 18 days, starts to kick in. The US cases total started up again on June 18. That puts the period of observing a renewed rise in deaths as scheduled for about June 6-8.
Texas ended its grace period 7 days ago - deaths have been rising sharply since June 25, and equaled our previous all-time one-day record on June 30.
All opinions expressed in this blog are solely those of the authors, and do not reflect those of University of Texas Rio Grande Valley, or any organizations of which either is a member.