[OC] Looking at total deaths by ANY cause, the USA is more than 250,000 deaths over budget for 2020. This graph shows how many deaths have happened more or less than expected each week since 2017.

Image from preview.redd.it and submitted by 120Macky
image showing [OC] Looking at total deaths by ANY cause, the USA is more than 250,000 deaths over budget for 2020. This graph shows how many deaths have happened more or less than expected each week since 2017.

120Macky on September 21st, 2020 at 02:36 UTC »

Data source: data.cdc.gov

Tools used: Python 3, Jupyter notebooks, plotly

The data compiled by the CDC is confirmed deaths, regardless of cause. As deaths can take weeks to be reported, the last week on the visual is the week ending August 15, 2020. The data for some weeks may increase, but not decrease.

"Expected deaths" for each week are the mean number of deaths for that week in 2017-2019. For instance, if there are 50,100 deaths in the first week of 2017, 50,500 in 2018, and 50,000 in 2019, the expected deaths for that week of the year would be 50,200. In this graph, the first bar would be -100, indicating 100 fewer deaths than expected for the first week in 2017. A linear regression would provide a more statistically informative expected number, but there is not sufficient data available from the CDC to make a such an estimate.

Edit: My Medium article on the visual

soodesne on September 21st, 2020 at 02:42 UTC »

Could you add a line of running total on a secondary axis?

If possible a version of graph only for 2020 with expected numbers as a line would give more readibility - right now it's pushed to right.

tadhgcube on September 21st, 2020 at 03:37 UTC »

Very interesting how one January is way over, and the next is way under... Any thoughts?