As I mentioned in Sunday's numbers update, I heard media reports that Arizona had a "spike" in new cases on Saturday the 19th. That didn't affect the averages I was posting for the past week, so I wondered what they were talking about.
It seems the origin of this was an AP wire story: "Arizona Department of Health Services officials reported 641 new cases and 16 deaths Saturday. It was the highest single-day case number since June 2." It also said Arizona reported 385 cases the next day. No hyperbole there, just a routine report on numbers from the state.
Then Joseph Choi at The Hill picked up this story and turned the number into a "spike" and implied a reversal of fortune for State 48:
Arizona recorded a small spike in coronavirus cases this past week when it recorded more than 600 new cases on Saturday, marking a sudden increase in what has been a downward trend in cases for the state. Cases in Arizona have been consistently dropping since spiking in January; however, The Associated Press reported that 641 cases were confirmed in Arizona on Saturday, the highest reported single-day increase in about a month.
To see how wrong/misleading these two sentences are, let's look at the numbers in context (numbers retrieved from CDC Socrata API):
First, normally, when people say some statistic (like unemployment) spikes, they mean it goes up and stays there for some time. What Choi calls a "small spike" I would call a "blip." Anyone (like AP) who bothered to look at the Sunday cases could see that's what it was, rather than some ominous change in disease rates.
More importantly it is part of normal variation. Looking at the chart we can see several similar blips. They occur at a suspiciously regular interval, leading me to believe they are due to some weekly variation in the reporting process.
That's why, as I have
said before, one should
always use averages when drawing conclusions about numbers like this. As you can see from the orange line on the chart, these blips have had a negligible effect on the average.
Second, it is not the case, as claimed by Choi, that Saturday's count was the highest in about a month. There was a much bigger number on June 2.
Finally, Choi's claim that the Arizona numbers "have been consistently dropping since spiking in January" is not accurate either. As you can see from the chart above, this was perhaps true until the beginning of June, but by about the first week of the month the average plateaus at around 400 cases a day. It is not consistently going down. It may even be trending up a slightly over this month.
This is
not the first time I have blogged about sloppy data journalism. The impulse to sensationalize a story should not overrule giving an accurate picture of the situation.
