SERVING THE QUANTITATIVE FINANCE COMMUNITY

  • 1
  • 4
  • 5
  • 6
  • 7
  • 8
  • 31
 
User avatar
Cuchulainn
Topic Author
Posts: 63413
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Models for Covid-19

March 24th, 2020, 7:08 pm

Why did you pick ALMEIDA? 
I've already answered those kinds of questions.. 

Suggest a model and post it.

This is what you wrote yesterday
Why pick that model? There are many models in this field! Either use the one they are using, or the best or the worst, for checking and comparison!

Most of these models probably don't work anyways because they have no optimal control and don't apply to COVID-19 (I guess my source is reasonably accurate).
My C++ Boost code gives
262537412640768743.999999999999250072597198185688879353856337336990862707537410378210647910118607313

http://www.datasimfinancial.com
http://www.datasim.nl
 
User avatar
Herd
Posts: 562
Joined: October 2nd, 2003, 12:48 pm

Re: Models for Covid-19

March 24th, 2020, 7:17 pm

According to the FT Some professor Gupta from Oxford has led a study saying that maybe half of the U.K. population got it and said “I am surprised that there has been such unqualified acceptance of the Inperial model”.

Couldn’t find the paper.
Google Oxford study coronavirus to find the FT article
 
User avatar
Cuchulainn
Topic Author
Posts: 63413
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Models for Covid-19

March 24th, 2020, 7:20 pm

According to the FT Some professor Gupta from Oxford has led a study saying that maybe half of the U.K. population got it and said “I am surprised that there has been such unqualified acceptance of the Inperial model”.

Couldn’t find the paper.
Google Oxford study coronavirus to find the FT article
This unqualified acceptance spurred me to start this thread. And still nothing beyond brainstorming.
Last edited by Cuchulainn on March 24th, 2020, 7:56 pm, edited 1 time in total.
My C++ Boost code gives
262537412640768743.999999999999250072597198185688879353856337336990862707537410378210647910118607313

http://www.datasimfinancial.com
http://www.datasim.nl
 
User avatar
Herd
Posts: 562
Joined: October 2nd, 2003, 12:48 pm

Re: Models for Covid-19

March 24th, 2020, 7:37 pm

What to do?

1. Check the numerical implementation?
2, Discuss assumptions that could be ok for flu but not for that virus in other words is that flu model appropriate for that virus)
3. Estimation of inputs
 
User avatar
Cuchulainn
Topic Author
Posts: 63413
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Models for Covid-19

March 24th, 2020, 7:55 pm

According to the FT Some professor Gupta from Oxford has led a study saying that maybe half of the U.K. population got it and said “I am surprised that there has been such unqualified acceptance of the Inperial model”.

Couldn’t find the paper.
Google Oxford study coronavirus to find the FT article
Where's the FDA when you need them?
My C++ Boost code gives
262537412640768743.999999999999250072597198185688879353856337336990862707537410378210647910118607313

http://www.datasimfinancial.com
http://www.datasim.nl
 
User avatar
Cuchulainn
Topic Author
Posts: 63413
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Models for Covid-19

March 24th, 2020, 7:57 pm

What to do?

1. Check the numerical implementation?
2, Discuss assumptions that could be ok for flu but not for that virus in other words is that flu model appropriate for that virus)
3. Estimation of inputs
0. Study the models.
My C++ Boost code gives
262537412640768743.999999999999250072597198185688879353856337336990862707537410378210647910118607313

http://www.datasimfinancial.com
http://www.datasim.nl
 
User avatar
Paul
Posts: 10922
Joined: July 20th, 2001, 3:28 pm

Re: Models for Covid-19

March 24th, 2020, 8:04 pm

Why did you pick ALMEIDA? 
I've already answered those kinds of questions.. 

Suggest a model and post it.

This is what you wrote yesterday
Why pick that model? There are many models in this field! Either use the one they are using, or the best or the worst, for checking and comparison!

Most of these models probably don't work anyways because they have no optimal control and don't apply to COVID-19 (I guess my source is reasonably accurate).
That was hours ago!

Herd is right.

Ignore flu. Ignore chicken pox.

Decide what is important for this problem and start from scratch. But keep it simple.

Independent variable is time. Forget anything spatial, that’s irrelevant now.

Dependent:
Deaths/Recovered
Infected
Hospital beds

Maybe age as parameter.
 
User avatar
Cuchulainn
Topic Author
Posts: 63413
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Models for Covid-19

March 24th, 2020, 8:08 pm

From the FT.

The modelling by Oxford’s Evolutionary Ecology of Infectious Disease group indicates that Covid-19 reached the UK by mid-January at the latest. Like many emerging infections, it spread invisibly for more than a month before the first transmissions within the UK were officially recorded at the end of February. The research presents a very different view of the epidemic to the modelling at Imperial College London, which has strongly influenced government policy. “I am surprised that there has been such unqualified acceptance of the Imperial model,” said Prof Gupta.  However, she was reluctant to criticise the government for shutting down the country to suppress viral spread, because the accuracy of the Oxford model has not yet been confirmed and, even if it is correct, social distancing will reduce the number of people becoming seriously ill and relieve severe pressure on the NHS during the peak of the epidemic. The Oxford study is based on a what is known as a “susceptibility-infected-recovered model” of Covid-19, built up from case and death reports from the UK and Italy. The researchers made what they regard as the most plausible assumptions about the behaviour of the virus. The modelling brings back into focus “herd immunity”, the idea that the virus will stop spreading when enough people have become resistant to it because they have already been infected. The government abandoned its unofficial herd immunity strategy — allowing controlled spread of infection — after its scientific advisers said this would swamp the National Health Service with critically ill patients. 
Last edited by Cuchulainn on March 24th, 2020, 8:09 pm, edited 1 time in total.
My C++ Boost code gives
262537412640768743.999999999999250072597198185688879353856337336990862707537410378210647910118607313

http://www.datasimfinancial.com
http://www.datasim.nl
 
User avatar
Cuchulainn
Topic Author
Posts: 63413
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Models for Covid-19

March 24th, 2020, 8:08 pm

Why did you pick ALMEIDA? 
I've already answered those kinds of questions.. 

Suggest a model and post it.

This is what you wrote yesterday
Why pick that model? There are many models in this field! Either use the one they are using, or the best or the worst, for checking and comparison!

Most of these models probably don't work anyways because they have no optimal control and don't apply to COVID-19 (I guess my source is reasonably accurate).
That was hours ago!

Herd is right.

Ignore flu. Ignore chicken pox.

Decide what is important for this problem and start from scratch.  But keep it simple.

Independent variable is time. Forget anything spatial, that’s irrelevant now.

Dependent:
Deaths/Recovered
Infected
Hospital beds

Maybe age as parameter.
Off you go. Good luck. I am taking a slightly different path.
Maybe read what the giants in this field have done. SEIR + control is a standard it seems and is used by the Chinese scientists. 

 Ignore chicken pox. 
No. It is the prototype to test my MSEIR code, And I have replicated varicella. A kind of sanity clause. Now we can add to the model.
My C++ Boost code gives
262537412640768743.999999999999250072597198185688879353856337336990862707537410378210647910118607313

http://www.datasimfinancial.com
http://www.datasim.nl
 
User avatar
Cuchulainn
Topic Author
Posts: 63413
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Models for Covid-19

March 24th, 2020, 8:23 pm

A message that someone sent (read carefully)

The Keeling model and some of the Chinese models use an SEIR compartmental structure with time varying parameters. This goes in the right direction but it should probably be split into hospitalized and quarantined when assessing economic impacts. Optimal control models are not new in epidemiology but were used already in the early 1970's and dynamic programming had been used as early as 1968 in epidemiology.
(BTW Matt Keeling is at Warwick).
Last edited by Cuchulainn on March 24th, 2020, 8:28 pm, edited 1 time in total.
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

March 24th, 2020, 8:27 pm

What use cases could be tested? BTW the metadata is in EXCEL  Image

I was always led to believe that ML/ANN only works when there is enough training data? So, how does it work for Covid-19?
yes apologies Dan, I wasn't exactly specific :) Referring to the dataset trackstar shared, it consists of many documents; I'm happy to provide a cloud service/ API for processing ~ 40k documents per day in order to create text from PDFs / keyword extraction for bag of words or similar NLP tasks. 

If we're more interested in the ODEs, then I have some optimized fortran that might be of use. I didn't look closely, but the system you shared is similar to the kernel of a batch reactor / CFD code I work with on a daily basis. It's richardson/  rosenbrock, but maybe this isn't a stiff system, I only glanced at a couple of parameters that looked very different in magnitude.

I think it might be interesting to work into the model the effect of bad actors / poor / missing information. I find the timing with the recent oil war curious, one of the countries involved seems awfully quiet about the little predicament that has the rest of the developed/ing world shtting the bed. I digress.

I guess I buy the scaling arguments about ventilators / hospitals. Tony stark has offered to ramp up production, and afaik the US created the national guard for times such as these. Maybe the models should take into account hardware scaling efforts, which as we all know is much more difficult than software.
 
User avatar
Cuchulainn
Topic Author
Posts: 63413
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Models for Covid-19

March 24th, 2020, 8:36 pm

Thanks, zeta, no problem. The ODE solver part in C++ (boost odeint) is all taken care of. Even good olde Cash Karp is excellent. The ODEs are not stiff and I use compile-time std::array<double, 4>.

I suppose fitting will be needed + more HPC crunching. I am only following leaders, Why, only two days ago I couldn't even spell epistomology.

According to IDA Ireland, 50% of ventilators used in acute hospitals worldwide are made in Ireland.
My C++ Boost code gives
262537412640768743.999999999999250072597198185688879353856337336990862707537410378210647910118607313

http://www.datasimfinancial.com
http://www.datasim.nl
 
User avatar
Paul
Posts: 10922
Joined: July 20th, 2001, 3:28 pm

Re: Models for Covid-19

March 24th, 2020, 9:04 pm

I'll probably start a separate thread for modelling issues as opposed to numerical. Meanwhile this is the world we live in now.
2020-03-24_110017 - Model 1.jpg
 
User avatar
zeta
Posts: 1973
Joined: September 27th, 2005, 3:25 pm
Location: Houston, TX
Contact:

Re: Models for Covid-19

March 24th, 2020, 9:22 pm

 lol nice sketch paul - the hospital needs the tesla logo on it tho.

My silicon valley experience was a lot like the tv show, but I did enjoy working on the hpc piece for a company scaling mri to whole population level. I admire tony stark, if anyone can get us out of this pickle, it's him.
 
User avatar
Paul
Posts: 10922
Joined: July 20th, 2001, 3:28 pm

Re: Models for Covid-19

March 24th, 2020, 9:23 pm

Him, or numerical analysts apparently!
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