Covid-19-data: Data Issue: NJ case counts 7-day moving average
Describe the issue:
Looking at NJ cases counts, I feel that 7 day average is reported about 30% higher than it should be based on daily stats. Same seems to happen for NY too.
Here is the page I am referring to:
https://www.nytimes.com/interactive/2020/us/new-jersey-coronavirus-cases.html
avarakin
All 2 comments
Hi @avarakin, I'm not sure what you mean by what it "should be". Perhaps you are simply using a different method for calculating the moving average?
albertsun
on 12 May 2020
I think the explanation is that you are intuitively doing a centered smoothing of the data, taking an average of data from 3 days before through 3 days after a given day. The Times plots are doing a non-centered "simple moving average" of the previous 7 days. This means that during a time when the cases are dropping each day, the 7-day average will be higher on average than the cases on a particular day. A centered smoothing method is essentially a more accurate method (and is more widely used in science and engineering) but it is bit harder to explain and it can't be directly applied to the most recent 3 data points, because that would require knowing the future. A simple moving average is more widely used in finance. For more details, see https://en.wikipedia.org/wiki/Moving_average.
You can check the curves by manually averaging 7 days of data and verify it is working as advertised as a simple moving average of the past 7 days.
My suggestion is that this issue can be marked as resolved.
gwhammett
on 30 May 2020
Related issues
harley164
路
5Comments
harrislapiroff
路
5Comments
brambow
路
5Comments
WeileiZeng
路
3Comments
bcunning
路
5Comments
Most helpful comment
I think the explanation is that you are intuitively doing a centered smoothing of the data, taking an average of data from 3 days before through 3 days after a given day. The Times plots are doing a non-centered "simple moving average" of the previous 7 days. This means that during a time when the cases are dropping each day, the 7-day average will be higher on average than the cases on a particular day. A centered smoothing method is essentially a more accurate method (and is more widely used in science and engineering) but it is bit harder to explain and it can't be directly applied to the most recent 3 data points, because that would require knowing the future. A simple moving average is more widely used in finance. For more details, see https://en.wikipedia.org/wiki/Moving_average.
You can check the curves by manually averaging 7 days of data and verify it is working as advertised as a simple moving average of the past 7 days.
My suggestion is that this issue can be marked as resolved.