I’m not a stats guy so I can’t vouch for the superiority of the method, but today I took my chart listing the daily count of newly confirmed COVID cases in New Hampshire and added a 5-day moving average line. Each point on the red line line of the chart above is the average of new cases over the past 5 days, which are plotted on the black line.
5-day moving averages are common in many systems that chart things on a daily basis, like stocks, in order to smooth out day-to-day fluctuations and better show patterns. How significant it is here, I don’t know, but there are certainly daily fluctuations in our new-case count. Some are probably due to anomalies in testing and reporting of tests (e.g., Monday numbers tend to go up because weekend tests are slow to report) rather than reflections of changes in actual caseload.
The one lesson from the red line? New cases are still edging up, on average; despite that hopeful looking leap and fall earlier this month, the peak has yet to come.
Return to the Concord Monitor
I think this a good way of tracking the trend but it is somewhat flawed because testing is just starting to ramp up. Love your columns.
Yes, but the real issue is the number of tests, or, the INCREASE in the number of tests. It is obvious that if you double the number of tests you are administrating, you will have an increase in the number of new cases. So, stating that the number of new cases is increasing is misleading. In my opinion, the percentage of new positive tests, or the number of hospitalizations might be a better indicator of the status of the population’s health.
See https://dnyuz.com/2020/04/17/why-epidemiologists-still-dont-know-the-death-rate-for-covid-19/
DO NOT USE 5-DAY MOVING AVERAGES!!!! There is a reason 5-day moving averages makes sense for stocks and other business transactions: There are 5 days in a business week. The 5-day moving average removes any weekly fluctuation. But, a 5-day moving average makes no sense for data that is reported 7 days a week. Look at the data. There is a clear pattern there, where weekend reporting is always decreased. That’s what a 7-day moving average is meant to abstract out. When you use a 5-day moving average, it is going to be cyclic, because you have not averaged out that weekly pattern, but are instead accentuating it.
And yes, I know the people at Johns Hopkins are using a 5-day moving average. The fact they are from JHU does not imply they know what they are doing…
If you look at more recent stories you’ll see I used 14-day moving averages. This was an early piece.