Coronavirus Curve in Context
Updated: a day ago
One morning while watching the news with my wife I said, “They need to add the moving average to that graph [daily cases of Coronavirus].”
“What?” she responded
“You know, the moving average so we can see the trend of the cases, and not worry about a single day spike ” I answered.
“Whatever…” she said.
This conversation was in the morning and three things occurred:
1. I didn’t illustrate my point
2. Statistics sound like a foreign language
3. She hadn’t had coffee yet
In this post, I'll apply two statistical tools to provide context to the Coronavirus Curves.
If I sound like a nerd, just look at the before and after of the pictures.
No spin or bias – This is what the damn Covid-19 Charts mean
Edward Deming once said, “In God We Trust, all others must bring data.” In this post I'm going to cut the BS out, and only share results, not my opinion. We are using two tools that can help us predict trends and determine if the virus is under control, statistically speaking. The first tool is a run chart with a moving average. The second tool is a control chart next we are going to look a control chart. These charts put data into context to make informed decisions.
Run charts show direction
A moving average on a run chart predicts the future. Above is America and three states' positive cases by day for the last two months (June and July). On the left we have the bar charts we have all seen, and if you look at this so what? However, if we convert this graph into a moving average the data is smoothed out developing a trend line. A moving average takes historical data points to smooth over the numbers and show a trend. The graph on the right shows the 7- and 14-day moving average. Different sources use different moving averages, but the smaller the moving average, the more impact a single day has.
Control Charts show performance
Control charts add another perspective to data to make informed decisions. Control charts display the average against user defined targets (specification limits) and statistical defined targets (control limits). A control chart helps eliminate the noise of daily changes to see if the process is "in control” based on statistical or user defined limits. So far as testing increases for the Coronavirus so do positives cases. As the year progressed a new phrase, Percent Positive, was introduced to monitor the number of cases with the increased testing. I want to be clear setting the target is political, and I’ll share two sources. The White House sets the target at 10% and deems any state over 10% percent positive in the “red zone.” The World Health Organization recommends 5% positive. In this example, we’re going to use 10%. The control limits are based on how much the numbers change from day to day (otherwise known as standard deviation), and in this example we are using a level 1 for control limits.
On the left we have the percent positive test results for the last 2 months.
If we overlay the average percent positive (black line) with the target (red line) with limits (yellow lines) we see the percent positive is above average and target. The control chart shows the results were never below the 2-month average or even close to the target in certain states, but well below in Vermont.
We are all in this together
This is not a political or public health post. Simply applying statistical charts to the data and curve we look at daily provides insight. In our personal and professional lives, statistics are all around us. Next time, remember to look at the moving average to spot a trend or specification limits to see how far you are off from the goal.
Communities can control the transmission of Covid-19
Don't worry about daily cases or headlines; look at the trend
Every case matters
Here's a link to the data source I used: Covid Data