This isn't J programming, yet; "chat" seems a bit too trivial, but
will do.
FYI - out today, I think. Based on Europe data only, sorry.
Reasonable hopeful results, though confidence intervals still wide. I
haven't encountered
50% credible intervals before.
https://www.imperial.ac.uk/media/imperial-college/medicine/sph/ide/gida-fellowships/Imperial-College-COVID19-Europe-estimates-and-NPI-impact-30-03-2020.pdf
I won't try attaching it!
Mike
-------- Forwarded Message --------
Subject: Re: [Jprogramming] COVID-19 and simulations in J
Date: Sat, 28 Mar 2020 19:28:56 +0100
From: Dimitri Georganas <[email protected]>
Reply-To: [email protected]
To: [email protected]
All sorted out - R_HOME was the problem (I mostly use jqt and it would quit
every time I ran Ropen'' without an error). This week I'll try to port
everything to J with R as server. Thanks.
On Sat, Mar 28, 2020 at 7:12 PM Raul Miller <[email protected]> wrote:
Also relevant here, is this:
https://github.com/vmchale/coronavirus
(I had mentioned this previously on the chat forum, but since it's
really J that might have been the wrong forum.)
((Also, if you're a J programmer who isn't familiar with the 'make'
program, you might need to run the wget commands from the Makefile
manually.))
FYI,
--
Raul
On Sat, Mar 28, 2020 at 7:18 AM Dimitri Georganas <[email protected]> wrote:
Hi,
I've been modeling the spread of the disease using a Gompertz function
and
its 1st, 2nd and 3d derivative to determine rate, acceleration and
'jerk',
but I couldn't get the R addon to work with my J installation, so I did
this part in R. (I couldn't find Levenberg Marquardt curve
fitting libraries/addons in J) Then I used the a,b,c I got from the R
code
in a J script to estimate the number of IC beds required, assuming a 10%
critical cases scenario. Well, Houston, we have a problem.
I could port the R code to J if I could either get the R addon to work or
find a Levenberg-Marquardt implementation in J.
IC impact code https://github.com/dimgeo/IC-covid
R code with Gompertz curve fitting and derivative
https://github.com/dimgeo/covid
Best regards,
Dimitri
On Fri, Mar 27, 2020 at 7:24 PM Robert Herman <[email protected]>
wrote:
I found this article:
Modeling the COVID-19 Outbreak with J
<
https://datakinds.github.io/2020/03/15/modeling-the-coronavirus-outbreak-with-j
I took the author's code, which is a modified SEIR model
(The_SEIR_model
<
https://en.wikipedia.org/wiki/Compartmental_models_in_epidemiology#The_SEIR_model
),
and updated some of the data and recalculated the variables. I then
pilfered some of Michal's (Tangentstorm, you rock!) code for animating
things in J from his talks. This is helping me to push my learning of
J,
and keep it interesting.
Two disclaimers: The author of the linked article doesn't claim to be
an
epidemiologist, and I am only doing this to learn J, and cannot
comment on
any of the validity of the algorithm or simulations it produces.
Now, on to my two issues (full code at end):
1. I know I have to figure out a way to have the* infect_display^:144
starting population* call feed the *update* loop, otherwise it is
trying to
run all 144 days of simulation in one call to update at a frequency
set by
*wd
'timer 600'*. I am working on this, and realize I somehow have to run
it
outside of update, and then feed update "a day at a time". The goal is
to
see the same person (grid location) evolve over the simulation. I know
this
is simple, but i haven't gotten there yet. Any tips will be
appreciated!
2. When I had initially set up the code to run 25 people instead of
144,
the color index I had setup with the DSEIR variable exhibit two strange
glitches. First, squares that were supposed to be 1=White, were
rendering
as 0=Black=Dead, and sometimes colors appeared "blended". An yellowish,
off-white for yellow, and at times bright yellow. When I went to 144
people
and it is trying to run in update, the grid appears to be correctly
indexing and not moving askew from frame to frame, but the colors are
inconsistent.
BTW, thanks to Linda's article in Journal of J - Vol.4, No.1 for the
RGB
list to then pull my desired colors from in the DSEIR variable (Dead,
Susceptible, Exposed, Infectious, Recovered):
*RGB=: (#:i.8) { 0 255*
*DSEIR =: (0 7 6 4 2 { RGB)*
Here is my integration of the two sources of code to try and animate
the
outbreak per the article, and thanks!:
*load 'viewmat'*
*coinsert'jgl2'*
*wd 'pc Covid-19-Sim closeok' NB. parent control (window) named 'w0'*
*wd 'minwh 500 500; cc g0 isidraw;' NB. add an 'isidraw' child control
named 'g0'*
*wd 'pshow; pmove 40 510 0 0' NB. show the window at the given
coordinates.*
*wd 'sm focus term' NB. session manager: bring terminal to front*
*wd 'psel Covid-19-Sim; ptop' NB. bring our window to front*
*vmcc =: viewmatcc_jviewmat_ NB. viewmat to a child control*
*step =: render @ update NB. each step, we'll call those two in
sequence*
*RGB=: (#:i.8) { 0 255*
*DSEIR =: (0 7 6 4 2 { RGB)*
*update =: verb define*
* im =: 12 12 $ infect_display^:144 starting_pop*
*)*
*render =: verb define*
* DSEIR vmcc im;'g0'*
* glpaint''*
*)*
*sys_timer_z_ =: step_base_*
*wd 'timer 600'*
*ppl =: 144 NB. How many people are in our simulation?*
*Beta =: 0.106271 NB. The rate at which susceptible people become
exposed.
RPH - dependent on R0 = 2.2 3/25/2020*
*Sigma =: 0.066967 NB. The rate at which exposed people become
infectious.
RPH - (1-sigma)^10=0.5, 10 is incubation period.*
*Gamma =: 0.056126 NB. The rate at which infectious people recover.
RPH -
adjusted to 12 days, so (1-Gamma)^12=0.5.*
*Xi =: 0.02284 NB. The rate at which recovered people lose immunity and
become susceptible. RPH - 30 days*
* NB. is partially based upon man on Diamond Princess who presented
symptoms around 30 days otherwise made up.*
*R =: 2.2 NB. How many people will 1 person infect? RPH - based on
latest
CDC estimate 3/25/2020*
*CFR =: 0.015 NB. Case Fatality Rate - RPH based upon 0.25% to 3% from
CDC,
so 1.5% is higher than average.*
*D =: 0.0011619 NB. Death rate per day while symptomatic - RPH based
upon
13 days to death as seen and per CDC.*
*ClosenessStdDev =: 5 % 3*
*ClosenessMean =: 1*
*rand_norm =: 3 : '(2&o. 2p1 * ? y) * %: _2 * ^. ? y'*
*closeness =: 4 %~ 2 %~ (+ |:) ClosenessMean + ClosenessStdDev *
rand_norm
(ppl,ppl) $ 0*
*risk =: closeness - closeness * =i.ppl *
*NB. risk =: (ppl,ppl)$0*
*NB. 0 = Dead*
*NB. 1 = Susceptible*
*NB. 2 = Exposed*
*NB. 3 = Infectious*
*NB. 4 = Recovered*
*starting_pop =: 2 , (ppl - 1) $ 1*
*infect =: 3 : 0*
* can_spread =. (1&< *. 4&>) y*
* susceptible =. y = 1*
* exposed =. y = 2*
* infectious =. y = 3*
* recovered =. y = 4*
* infectiousness =. can_spread ,./ . * risk*
* susceptible_to_exposed =. (Beta * susceptible) * (+/ can_spread) %~
+/ |:
susceptible * infectiousness*
* exposed_to_infectious =. Sigma * exposed*
* infectious_to_recovered =. Gamma * infectious*
* infectious_to_dead =. D * infectious*
* recovered_to_susceptible =. Xi * recovered*
* 1 (I.recovered_to_susceptible>?ppl$0)} 4
(I.infectious_to_recovered>?ppl$0)} 0 (I.infectious_to_dead>?ppl$0)} 3
(I.exposed_to_infectious>?ppl$0)} 2 (I.susceptible_to_exposed>?ppl$0)}
y*
*)*
*infect_display =: 3 : 0*
* out =. infect y*
* smoutput out*
* smoutput
('Susceptible';'Exposed';'Infectious';'Recovered';'Dead'),:(+/1=out);(+/2=out);(+/3=out);(+/4=out);(+/0=out)*
* out*
*)*
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
--
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
