Sunday, April 5, 2020

State By State Coronavirus Modeling

I have previously written about the public health modeling for coronavirus. While I believe it leaves a great deal to be desired compared to the models used by meteorologists, they probably have value in a comparative sense. To test that hypothesis, I want to call to your attention a model that allows state-by-state models. The model is here.

Here in Kansas, where the lockdown was both relatively early and comprehensive, the graph looks like this.
It projects a total of 640 deaths by August 1with almost all of the deaths occurring by about May 20. Dividing by population, .0219 of the citizens of the state are forecast to pass away.

Kansas can be compared with Alabama where protective measures were put in place later and not as comprehensively.
The model projects 5,516 deaths or .112% of the population. That percentage is a huge difference -- more than 5 times the rate of deaths -- as compared to Kansas. Let's pray that the deaths are far fewer than projected.

My objection to these models is the confidence spread (tinting). In Kansas, the June 1st confidence interval ranges from 220 to 1,225. That is a huge spread that, in my opinion, is unacceptably large. Once all of this is over, it is vital that public health authorities create better models for future disease outbreaks.

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.