Sorry it has been a few days. My readers are texting me saying, GET YOUR NEXT POST UP!
I had a post ready to go last week and somehow it didn’t get posted, so today I’ll compare that model’s forecast to what happened.
But first, it has been a while since I ranted about King County Public Health. This week, we (again) saw a huge upward revision in previous months’ data. For example, The May 10th number was revised up by 10. Approximately a 10% increase. Which begs the question, why is May data off by as much as 10%?
In my head, I’m picturing these dusty boxes filled with COVID tests that the agency put in the corner to deal with later. They’re only now getting around to the results and updating the data. Do you think the people were notified that they had a positive test?
To top it off, there is no data available for last Thursday. This is the first time that King County has put up a goose egg. No tests reported. No infections. Pathetic.
And to think there are major public policy decisions being based on this data.
So now, to the models and the results:
LAST WEEK: Numbers were trending in the right direction in King County. The 2-week average was trending downward at a clip that will reach 0 in 8-10 weeks.
THIS WEEK’S UPDATE: The 2-week trend continues. About 7 weeks to get to 0.
LAST WEEK: The polynomial regression model continues to paint a more rosy picture (with better fit). The model continues to show a steep drop in infections is just around the bend.
THIS WEEK’S UPDATE: The model from last week didn’t come to pass. The model is sticking firm with a steep decline “soon.”
However, you’ll notice there was a big drop in the model’s power. Basically, the model isn’t as certain about what will happen. In other words, it looks like model is about to break again. I’m going to need to come up with something else. (Or, maybe the numbers will change).
And it’s hard to know the drivers of this due to the fact that old data keeps changing. When data back to April and May is changing at the end of August, is that the driver of the model being less predictive. Is the data even accurate?
Last point: Deaths continue to be way down. Only a handful over the past few weeks, which is great. And, the CDC recently came out and said that 94% of deaths were in people that had underlying health issues. With the recent surge in cases yet lack of hospitalizations and deaths, you have to consider that COVID-19 might be a mild disease, similar to the cold or flu. As a part-time economist, I ask, Is all this lockdown for nothing? What is the ROI of the lockdown?