SERVING THE QUANTITATIVE FINANCE COMMUNITY

 
User avatar
Cuchulainn
Topic Author
Posts: 62370
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Models for Covid-19

April 28th, 2020, 9:27 pm

The Singapore model is being quoted in the media now. Would be nice to see confirmation of their analysis, e.g. for UK.

You can get the same sort of plot as they get if you scale everything up (infected, susceptibles, parameters) and assume that the vast number of people are asymptomatic or not bothering reporting the illness. But I'm still getting that the best fit has currently more infected than the initial population of susceptibles! I.e. fails the most basic sanity clause check. Yes, Cuch, there's no sanity clause!
So, the party of the first party is out of his mind? You see them sometimes walking around blindly with dead eyes, following orders, not knowing what they do, not caring.”
I used to be a requirements analyst in the old days.  These things surface during meetings. 
 
User avatar
katastrofa
Posts: 9319
Joined: August 16th, 2007, 5:36 am
Location: Alpha Centauri

Re: Models for Covid-19

April 28th, 2020, 10:58 pm

Sanity Clause is in December... now we have Bank Holidays moved to Friday to coincide with the anniversary of you know what.

BTW, the model for Singapore is an agent-based simulation. They re-calibrated some older model (which is again based on an even older model):
We used FluTE, an agent-based influenza epidemic simulation model, which accounts for demography, host movement, and social contact rates in workplaces, schools, and homes to estimate the likelihood of human-to-human transmission of SARS-CoV-2 should local containment fail.

Ferguson's simulation is based on the same technique, but it seems less detailed with regards to the population variables.
 
User avatar
Paul
Posts: 10787
Joined: July 20th, 2001, 3:28 pm

Re: Models for Covid-19

April 28th, 2020, 11:41 pm

Is this the same Singapore model? The one I see (referenced in the media) says this: "SIR (susceptible-infected-recovered) model is regressed with data from different countries to estimate the pandemic life cycle curves and predict when the pandemic might end in respective countries and the world, with codes from Milan Batista and data from Our World in Data." And Milan Batista code says "The function fitVirusCV19 implements the susceptible-infected-removed (SIR) epidemic model for the estimation of epidemy evaluation" and "The parameters of the model are obtained by minimization of the objective function, which is the sum of squares for residuals of values and sum of squares for residuals of values differences."
 
User avatar
Paul
Posts: 10787
Joined: July 20th, 2001, 3:28 pm

Re: Models for Covid-19

April 29th, 2020, 3:03 am

Now with added removal rate of infectives!

Almost identical fit to theirs. Total susceptibles: 1.9million. 


SingUKfitted2.png

The 64 million person question is...is that number of susceptibles plausible? 

Well...

I get a ratio of 38:1 for numbers actually infected to reported infected. So that's actually consistent if you scale up the Singapore SIR model and parameters!

Which suggests, and here I'm just eyeballing the charts, a total number of deaths of 30-40,000. (And that's adding in a guess for numbers that have not been reported previously.)

How much of this is affected by lockdown is anybody's guess!
 
User avatar
Cuchulainn
Topic Author
Posts: 62370
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Models for Covid-19

April 29th, 2020, 11:32 am

 FluTE, an agent-based influenza epidemic simulation 

What is an agent, precisely? Same as this which was hot about 20 years ago (esp mobile agents, very difficult/underdeveloped s/w area).
Or is it just a name?
 
User avatar
Cuchulainn
Topic Author
Posts: 62370
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Models for Covid-19

April 29th, 2020, 11:46 am

Is this the same Singapore model? The one I see (referenced in the media) says this: "SIR (susceptible-infected-recovered) model is regressed with data from different countries to estimate the pandemic life cycle curves and predict when the pandemic might end in respective countries and the world, with codes from Milan Batista and data from Our World in Data." And Milan Batista code says "The function fitVirusCV19 implements the susceptible-infected-removed (SIR) epidemic model for the estimation of epidemy evaluation" and "The parameters of the model are obtained by minimization of the objective function, which is the sum of squares for residuals of values and sum of squares for residuals of values differences."
Local or global minimum? Good old sum of squares.
 
User avatar
katastrofa
Posts: 9319
Joined: August 16th, 2007, 5:36 am
Location: Alpha Centauri

Re: Models for Covid-19

April 29th, 2020, 4:30 pm

Paul, it indeed sounds like there's more than one Singaporean model (I referred to this one: https://www.thelancet.com/journals/laninf/article/PIIS1473-3099(20)30162-6/fulltext)
It's quite funny. I'm waiting for the official announcement which model was right ;-)

Cuchulainn, an agent is a unit in the microsimulation - I've done people, computers, and plants so far, but economic models have companies, organisations, any other entities. You usually start to talk about agents when the microsimulation (which in the original "static" setting amounted to analysing or changing proportions of different groups in the population) has elements of agent-based modelling (interactions between agents, dependencies, some autonomy, etc.).
 
User avatar
Paul
Posts: 10787
Joined: July 20th, 2001, 3:28 pm

Re: Models for Covid-19

April 29th, 2020, 4:55 pm

Paul, it indeed sounds like there's more than one Singaporean model (I referred to this one: https://www.thelancet.com/journals/laninf/article/PIIS1473-3099(20)30162-6/fulltext)
It's quite funny. I'm waiting for the official announcement which model was right ;-)
Don't hold your breath! (Unless you are not wearing a mask, of course!)
 
User avatar
Paul
Posts: 10787
Joined: July 20th, 2001, 3:28 pm

Re: Models for Covid-19

May 4th, 2020, 5:42 pm

When it's all over and the postmortem has been conducted, we shall find that there is little difference between countries' numbers and strategies. Differences will be down in a large part to either reporting or something external such as climate. Reporting of deaths in different countries and the testing will be criticized. Some countries will count everything as Covid-19, some will get dates wrong, some will have tested many people and found the virus everywhere, where you die will affect whether you appear on a list, date of death and reporting of death will be seen as confusing. And some large countries with unpleasant politics will be found to have lied a lot. 

Here's a plot of deaths in NYC. It shows how many died on each day, and how that number changed a few days later. So this takes out one source of confusion, out of many, and is quite reassuring.

NYC1.png
 
User avatar
Cuchulainn
Topic Author
Posts: 62370
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Models for Covid-19

May 5th, 2020, 7:54 pm

Exclusive: Government scientist Neil Ferguson resigns after breaking lockdown rules to meet his married lover

https://www.telegraph.co.uk/news/2020/0 ... -breaking/

The epidemiologist leads the team at Imperial College London that produced the computer-modelled research that led to the national lockdown, which claimed that more than 500,000 Britons would die without the measures.

A pressing question is who will maintain the code base?
 
User avatar
Cuchulainn
Topic Author
Posts: 62370
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Models for Covid-19

May 7th, 2020, 7:58 pm

All this hullabaloo about C models, what?

The Yanks use Excel ..

Trump’s advisers released a ‘beyond stupid’ mathematical model of coronavirus deaths created in Excel by a controversial economist

https://www.businessinsider.nl/trumps-coronavirus-team-relies-on-stupid-model-by-controversial-economist-2020-5?international=true&r=US

Earlier this week, The Washington Post reported that the White House is relying in part on a so-called “cubic model” devised by Kevin Hassett, a top economic adviser, that shows COVID-19 deaths plummeting to zero by mid-May.

To better visualize observed data, we also continually update a curve-fitting exercise to summarize COVID-19's observed trajectory. Particularly with irregular data, curve fitting can improve data visualization

On Wednesday, Hassett told The New York Times he used “just a canned function in Excel, a cubic polynomial” to create the model, confirming theories circulating online.
Last edited by Cuchulainn on May 7th, 2020, 8:15 pm, edited 3 times in total.
 
User avatar
Cuchulainn
Topic Author
Posts: 62370
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Models for Covid-19

May 7th, 2020, 8:01 pm

Earlier this week, The Washington Post reported that the White House is relying in part on a so-called “cubic model” devised by Kevin Hassett, a top economic adviser, that shows COVID-19 deaths plummeting to zero by mid-May.

This one takes the biscuit; don't tell me, don't tell me, are they using cubic splines (on sparse data). If yes, then it's the pits.

To better visualize observed data, we also continually update a curve-fitting exercise to summarize COVID-19's observed trajectory. Particularly with irregular data, curve fitting can improve data visualization. A
 
User avatar
Paul
Posts: 10787
Joined: July 20th, 2001, 3:28 pm

Re: Models for Covid-19

May 7th, 2020, 8:08 pm

Excellent! I said that these models all give the same shapes so might as well cut out the middleman!
 
User avatar
Cuchulainn
Topic Author
Posts: 62370
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Models for Covid-19

May 7th, 2020, 8:24 pm

Excellent! I said that these models all give the same shapes so might as well cut out the middleman!
Yeah, they all use Excel cubic splines, not work for irregular data. Any IR quant/trader knows that. Andrea Germani solve it before lunch.

Some are born great, some achieve greatness, and some have Excel thrust upon them.
 
User avatar
Cuchulainn
Topic Author
Posts: 62370
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Models for Covid-19

May 7th, 2020, 8:29 pm

Tomas Philipson, Trump’s current CEA chair, tweeted back at Furman, attacking his credentials and accusing him and Krugman of not understanding that the modeling was “data smoothing,” rather than “forecasting.”

Image
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