I went back to the model (available here) today to check a piece of information and was shocked by what I found. The model has been re-run. In just two days, the number of COVID-19 deaths forecast to occur in Kansas has been dropped from 640 to 265 -- a tremendous reduction. I double checked, the model's initial parameters are the same, so it is apples-to-apples. The initial parameters are at the very top of the Kansas display.
![]() |
UW's model projection for Kansas published today |
While this is certainly good news in the sense that fewer people than forecast are going to lose their lives, it is demonstrating these models are useless as decision tools.
Why do I say that?
Here's where we began: A late-January forecast of 1-2 million deaths in the USA.
![]() |
Axios |
That is a reduction of 94% in forecasted deaths (from 1.5 million, the mean number of the initial forecast).
As I have previously written, these models are useless as decision tools. When the history of CV in the U.S. has been written, we will learn that the use of these models will have done a great deal of harm. And, I think it is fair to state that our reliance on experts, as opposed to carefully and objectively looking at the raw day-to-day numbers, has not served us well.
(c) 2020, Mike Smith Enterprises, LLC
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.