SERVING THE QUANTITATIVE FINANCE COMMUNITY

 
User avatar
Paul
Topic Author
Posts: 10915
Joined: July 20th, 2001, 3:28 pm

Re: Models for Covid-19 - analytics

April 10th, 2020, 5:16 pm

Also this: "The continuum SIR model does not accurately represent human contacts, especially when they are limited by social distancing. It is better to consider a network, although it is not clear what the correct network should be."
 
User avatar
zeta
Posts: 1973
Joined: September 27th, 2005, 3:25 pm
Location: Houston, TX
Contact:

Re: Models for Covid-19 - analytics

April 10th, 2020, 6:14 pm

I thought that was a keen insight too. new kind of science aside, the guy can be pretty astute
 
User avatar
Paul
Topic Author
Posts: 10915
Joined: July 20th, 2001, 3:28 pm

Re: Models for Covid-19 - analytics

April 10th, 2020, 6:53 pm

Good to see: "...it is not clear what the correct network should be."!
 
User avatar
Cuchulainn
Posts: 63379
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Models for Covid-19 - analytics

April 10th, 2020, 7:21 pm

I'm not looking for insights (do it in yers free time) ; I want answers, NOW!
Happy Easter.

correct network should be.
A big network. A bit like a porous medium?


The current ODEs (high school stuff?) are everywhere and nowhere ... what about ODEs on random graphs (Erdős–Rényi )

https://www.sciencedirect.com/science/a ... 6514000334

Then you can ask the usual questions that you usually ask of networks. e.g. 

. how did it get from Wuhan to NYC?
. shortest path from a to b
. how many connected components

Kevin Bacon 2.0

If you are at a loose end this weekend, Boost Graph is waiting

https://www.boost.org/doc/libs/1_72_0/l ... index.html
My C++ Boost code gives
262537412640768743.999999999999250072597198185688879353856337336990862707537410378210647910118607313

http://www.datasimfinancial.com
http://www.datasim.nl
 
User avatar
zeta
Posts: 1973
Joined: September 27th, 2005, 3:25 pm
Location: Houston, TX
Contact:

Re: Models for Covid-19 - analytics

April 10th, 2020, 8:48 pm

lol and a happy easter to you to Dan

I think the currently accepted hypothesis is that there was a migration of covid from Europe to NYC, much like after WW1 where returning serviceman brought the flu to philly. 

Wolfram has an interesting graph in the same notebook based on data from Singapore.
 
User avatar
Alan
Posts: 10468
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

Re: Models for Covid-19 - analytics

April 11th, 2020, 2:41 pm

Wolfram fans might like his analysis from a stream last month : https://www.wolframcloud.com/obj/s.wolfram/Published/COVID-19-Livestream-March-24.nb

The last plot is particularly telling ..
Very nice -- Thanks!

For everybody, the latest Economist has an interesting piece: 
Why a study showing that covid-19 is everywhere is good news
If millions of people were infected weeks ago without dying, the virus must be less deadly than official data suggest

Here is the study they refer to, which can be downloaded. For the modelers, it has a calibrated SEIR model. From the paper:

Introduction:
The ongoing SARS-CoV-2 pandemic continues to cause tremendous morbidity and mortality around the world [1, 2]. Regional preparation for the pandemic requires forecasting the growth rate of the epidemic, the timing of the peak, the demand for hospital resources, and the degree to which current policies may curtail the epidemic, all of which benefit from accurate estimates of the true prevalence of the virus within a population [3]. Confirmed cases are thought to be underestimates of true prevalence due to some unknown combination of patients not reporting for testing, testing not being conducted, and false-negative test results. Estimating the true prevalence informs the scale of upcoming hospital, ICU and ventilator surges, the proportion of individuals who are susceptible to contracting the disease, and estimates of key epidemiological parameters such as the epidemic growth rate and the fraction of infections which are sub-clinical.

The current literature suggests that the predominant symptoms associated with COVID are fever, cough and sore-throat; that is, patients often present with an influenza-like illness (ILI) yet test negative for influenza [4, 5]. With many COVID patients having a similar presentation as patients with influenza, existing surveillance networks in place for tracking influenza could be used to help track COVID.

Here, we quantify background levels of non-influenza ILI over the past 10 years and identify a recent surge of non-influenza ILI starting the first week of March, 2020. This surge of excess ILI correlates with known patterns of SARS-CoV-2 spread across states within the US, suggesting the surge is unlikely to be due to other endemic respiratory pathogens, yet is orders of magnitude larger than the number of confirmed COVID cases reported. Together this suggests that the true prevalence of SARS-CoV-2 within the US is much larger than currently appreciated and that even the highest symptomatic case detection rates are likely lower than 3% corresponding to approximately 9 million new ILI cases due to SARS-CoV-2. Our analysis provides empirical corroboration of previous hypotheses of substantial undocumented cases[6] yet places the estimated undocumented case rate an order of magnitude higher than prior reports [6]. Moreover, these updated prevalence estimates predict that epidemic doubling times greater than 3.5 days [7, 8] would be unable to account for the magnitude of the ILI surge. We test our hypothesis of sub 3-day doubling times in the US by analyzing both state and national COVID surveillance data, finding a broad agreement of doubling time less than 3.5 days in both confirmed COVID case counts and documented COVID deaths. Our findings highlight probable trajectories of the US epidemic and provide evidence for a conceptual model for COVID spread in the US in which more rapid spread than previously reported is coupled with a larger undiagnosed population to give rise to currently observed trends.
 
User avatar
Alan
Posts: 10468
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

Re: Models for Covid-19 - analytics

April 11th, 2020, 8:41 pm

 
User avatar
Paul
Topic Author
Posts: 10915
Joined: July 20th, 2001, 3:28 pm

Re: Models for Covid-19 - analytics

April 11th, 2020, 9:14 pm

You mean if a denominator is bigger then the fraction is smaller? This pandemic is teaching us so much mathematics.
 
User avatar
Alan
Posts: 10468
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

Re: Models for Covid-19 - analytics

April 11th, 2020, 9:46 pm

Yes, I think it was Wigner who first commented on the unreasonable effectiveness of mathematics in the natural sciences.  :D
 
User avatar
Alan
Posts: 10468
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

Re: Models for Covid-19 - analytics

July 21st, 2020, 6:40 pm

New metastudy cited in

todays (July 21, 2020) WSJ: 
How Deadly is Covid-19? Researchers are Getting Closer to an Answer

Result: Infection fatality rate estimated at 0.68%  (0.53-0.82%)

In the US, given the current cumulative death count, and taking into account the lag from infection to death, this translates into an estimate of  4.9-7.6% of the population infected  -- let's say as of a month ago. (This is my figure, not the authors). Would be helpful to have this confirmed or refuted by some large scale random tests.
 
User avatar
katastrofa
Posts: 9707
Joined: August 16th, 2007, 5:36 am
Location: Alpha Centauri

Re: Models for Covid-19 - analytics

July 21st, 2020, 10:04 pm

Why no-one can ever recover from COVID-19 in England – a statistical anomaly

What if PHE made the "error" to protect us from our government trying to improve the UK's covid death rate in international statistics?
 
User avatar
ISayMoo
Posts: 2368
Joined: September 30th, 2015, 8:30 pm

Re: Models for Covid-19 - analytics

July 22nd, 2020, 10:00 am

It's hopeless, our government is not even able to count the number of people dying, let alone make them die less often. I'm sure there's going to be an inquiry ;-)
 
User avatar
Cuchulainn
Posts: 63379
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Models for Covid-19 - analytics

July 22nd, 2020, 10:31 am

Professor _iwouldneedtolookatmynotes_ Chris Whitty on Newsnight did not know whether lockdown was 16 March or 23 March while Sir Patrick Vallance at that time was into herd immunity. 
Some precise dates are in my report and Richard Horton's book.

// Newsnight begins too late in the evening.
My C++ Boost code gives
262537412640768743.999999999999250072597198185688879353856337336990862707537410378210647910118607313

http://www.datasimfinancial.com
http://www.datasim.nl
 
User avatar
Cuchulainn
Posts: 63379
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Models for Covid-19 - analytics

July 22nd, 2020, 7:08 pm

It's hopeless, our government is not even able to count the number of people dying, let alone make them die less often. I'm sure there's going to be an inquiry ;-)
False opinions are like false money, struck first of all by guilty men and thereafter circulated by honest people who perpetuate the crime without knowing what they are doing.

Joseph de Maistre
My C++ Boost code gives
262537412640768743.999999999999250072597198185688879353856337336990862707537410378210647910118607313

http://www.datasimfinancial.com
http://www.datasim.nl
 
User avatar
Alan
Posts: 10468
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

Re: Models for Covid-19 - analytics

July 23rd, 2020, 9:54 pm

Another perspective on the current COVID situation in the US
(source: cdc)
WeeklyExcessDeaths.png
ABOUT WILMOTT

PW by JB

Wilmott.com has been "Serving the Quantitative Finance Community" since 2001. Continued...


Twitter LinkedIn Instagram

JOBS BOARD

JOBS BOARD

Looking for a quant job, risk, algo trading,...? Browse jobs here...


GZIP: On