Now are they wildly pessimistic?
The increasingly implausible IHME model with which we’re all familiar imagines deaths in the U.S. dropping below 1,000 per day just three days from now. Yesterday we registered 2,700. Today, according to the latest data, we’re slightly north of 2,000, which is worse than the 1,700-1,800 we were registering on weekdays last week. If you believe IHME, we should see daily deaths nationwide sink below 500 on May 10, then below 100 on May 25, and then reach single digits on June 13 before dipping to zero on June 30. That seems wildly unlikely.
These dang predictive models have sure done a lousy job predicting. And I don't know about you, but I find that frustrating. Are we gonna drop to near zero, have a slow but constant burn for months with occasional spikes, or just get another hump? Dunno. Ask the experts? They appear to be guessing and doing about as well as me.
Also, that linked pessimistic article is Allahpundit, so assign credibility based on that as you may.
4 comments:
T-bolt,
Its going to be a long slow burn. the problem with predictive models is
the entry assumptions. That compounded, as we see the model we are altering
our behavior so the initial assumptions are changing. That iterative
alteration means the models error is increasing.
However no model had the bug going to zero that's just so much fantasy.
It the whole thing was slow the burn so that people getting sick have a
hospital with a bed and all the needed treatment available.
That means the slow burn as a result as it never goes away until:
We fid out if it confers a immunity, and so far at best a maybe.
A vaccine.
Then we go for maybe a year or more till enough get to enough vaccine
to tamp the fire.
Eck!
For any model to be correct, you need two thing:
Accurate and complete input data,
And a thorough understanding of the underlying processes.
If you don't have those (and we don't), your "model" is more correctly called a SWAG.
This holds as true for the virus as it does for climate change.
Mathematical models usually include an assumption that is on the order of Assume a Spherical Cow of Uniform Density in a Frictionless Vacuum>.
Which obviously has some problems.
When it comes to people, there are more problems.
And then there is the known-unknowns. Early on we didn't have a good grasp of the accuracy of the tests. (rates for false-positive, etc.) And then there were loads of bad test equipment/bad reagent supplied from various sources.
Tough to build a model under the best conditions. Under those conditions...
Then when there became financial incentives for declaring a hospitalization/use of respirator as COVID-19 related, the figures around cause-of-death became very suspect.
Turns out that the best thing to look at may be the change in mortality from expected mortality. Actuaries have been studying that for good long time and have it nailed. See Mortality with Meep: Excess Deaths And Coronavirus "It is very tough to fudge the count of total dead people."
Post a Comment