This just in via Josh Thorpe.
I've been wondering if anyone had a methodology for calculating the
COVID19 R /Replication Number./
https://rt.live/
apparently adapted from our own (LANL) Luis Bettencourt's work on H5N1
using a Bayesian approach
https://github.com/k-sys/covid-19/blob/master/Realtime%20R0.ipynb
This notebook is pretty big and pretty dense, I'm trying to skim through
it and get a feel for it.
It appears superficially that they are using nothing more than reported
new cases smoothed by a Gaussian filter to remove reporting/test/delay
artifacts.
What I'm not clear on quite yet is how (if) this approach handles the
intrinsic delay between exposure and onset of symptoms sufficient to
yield a confirmed case? If that (variable) delay is not factored in
then the R(t) would seem to be a smeared reflection of R(t-n...t) where
"n" is the maximum number of days between exposure and confirmation.
I'll keep looking.
I wish the summary view had a time-slider to watch the states R(t)
evolve... the buttons included for different previous times (yesterday,
last week, 2 weeks, 3 weeks) give a hint of this.
I'm surprised at how high some of the R values were even over 4.0 for
some states at some time.
I'm also surprised at how many states seem to have dropped to/below
1.0. And also how many seem to have dipped below 1.0 and bounced back
up. This would seem to imply that many states hit a high level of
social distancing/hygiene and then relaxed it (recently?).
I also haven't sussed out why the different states have such different
error envelopes...
I look forward to others possibly digging into this and sharing their
observations.
- STeve
.-. .- -. -.. --- -- -..-. -.. --- - ... -..-. .- -. -.. -..-. -.. .- ... ....
. ...
FRIAM Applied Complexity Group listserv
Zoom Fridays 9:30a-12p Mtn GMT-6 bit.ly/virtualfriam
unsubscribe http://redfish.com/mailman/listinfo/friam_redfish.com
archives: http://friam.471366.n2.nabble.com/
FRIAM-COMIC http://friam-comic.blogspot.com/
