search

CSSEGISandData / COVID-19

Novel Coronavirus (COVID-19) Cases, provided by JHU CSSE
https://systems.jhu.edu/research/public-health/ncov/
29.11k stars 18.39k forks source link

time_series_covid19_deaths_US.csv errors #2231

Open methurst opened 4 years ago

methurst commented 4 years ago

Data for the last 3 days is incorrect (repeated) for several states, e.g. Massachusetts, Kentucky, Tennessee

jrmatchett commented 4 years ago

Something sure is amiss. I take a daily snapshot of the dashboard values at ~midnight Eastern time, and according to the current value there was an increase of 4,884 deaths over the past 24 hours, which is over 2,500 deaths greater than any other past 24-hour period.

methurst commented 4 years ago

USA death totals as reported by state/terr govts from 4/13-4/15 were 1450, 2299, and 2492, respectively. If JHU are using different sources/cut-off times, your totals could vary, but at least where state govts are reporting increases in daily deaths, the JHU numbers should not be flat.

methurst commented 4 years ago

@jrmatchett Error in USA deaths for 2020/04/16 still has not been corrected. JHU reports 4591 deaths, while the total reported by state/terr govts is 2136. In my opinion, JHU needs to invest some serious resources in validating the accuracy of the data you are reporting, even if that means reporting less data. As it is the quality is becoming so poor that it is in danger of becoming completely useless.

RetiredActuary commented 4 years ago

The MA data across much of April looks like someone tried to play catch-up after missing multiple days. I've also been collecting the snapshots & so I have that kind of view, but MA also just started publishing historical tables with counts by date of death, so compiling isn't even necessary anymore. Wish I could go in & fix it myself.