SERVING THE QUANTITATIVE FINANCE COMMUNITY

 
User avatar
Collector
Topic Author
Posts: 4734
Joined: August 21st, 2001, 12:37 pm

Re: How good or bad are the pandemic warning models?

February 13th, 2020, 12:08 pm

We will all be like Dr Rieux soon.



Image
"Thus the first thing that plague brought to our town was exile."

"Moreover, the epidemic seemed to be on the wane; on some days only ten or so deaths were notified. Then, all of a sudden, the figure shot up again, vertically. "

"The word “plague” had just been uttered for the first time. At this stage of the narrative, with Dr. Bernard Rieux standing at his window, the narrator may, perhaps, be allowed to justify the doctor’s uncertainty and surprise – since, with very slight differences, his reaction was the same as that of the majority of the townsfolk. Everybody knows that pestilences have a way of recurring in the world; yet somehow we find it hard to believe in ones that crash down on our heads from a blue sky. There have been as many plagues as wars in history; yet always plagues and wars take people equally by surprise."
 
User avatar
trackstar
Posts: 27504
Joined: August 28th, 2008, 1:53 pm

Re: How good or bad are the pandemic warning models?

February 13th, 2020, 12:32 pm

Someone should refit the curves backward (hypothetically, of course) since the vertical spike is at least partly due to a change in the methodology of detection in one province (Hubei) in China.

See (print) - Jump in Coronavirus Cases - Bloomberg Feb 13

Audio:


Fine to have and use a new form of data, but the whole curve should be adjusted upward, rippling across proportionally from right to left, at least as a specially marked retroactive “projection” and noting Hubei as the source of the change.

Also, other places should either adopt this methodology or Hubei should be listed separately.

Rough ideas (I leave the details to the tailors and experts here!):

a) Detection in Hubei before and after the change
b) Detection elsewhere before and after the change (if made efficiently)
or
c) Hypothetical increases elsewhere per a factor ~% change in a)

Apples and oranges,  C-viruses and bananas here.

But obviously, no matter how you may try to make appropriate adjustments, this is getting very messy.

What would be more alarming than this perhaps one-time spike in the stats would be evidence of an increase in R0, an even longer incubation time (14-24 days is bad enough), or findings that every one who has been on a plane from China is now a carrier. (I am being darkly humorous and dramatic on the last one, but 30,000 carriers running around in the US alone would not be too funny). The fact that you can be a carrier and asymptomatic or it only seems to be a mild cold is also a problem.

There are also some outside-the-health-care-system "labs" now, unfortunately for the people in them, but as models for transmission in unprotected environments, we have the apartment buildings, with recirculated air, and the cruise ships, which mean even closer confinement, probably no masks of any kind, and dependency on centralized kitchens and ship staff for food and water.

The Diamond Princess is not looking so sparkly right now. But the data coming out may be helpful.

Off to my erg dungeon. 6K today! Take care, folks. : )
 
User avatar
Paul
Posts: 10796
Joined: July 20th, 2001, 3:28 pm

Re: How good or bad are the pandemic warning models?

February 13th, 2020, 4:19 pm

These models tend to have similar shapes. S shaped for total number of deaths, i.e. a peak in death rate. But which is the right model since they all look the same?

Until yesterday I had an incredible fit for a simple model based on tw's "I wonder if one can solve the basic model with simple time-dependent parameters (R0 particularly) to describe improving policy responses to the the epidemic to see if there is some critical level you need to get to (and how fast) to nip it in bud." I didn't use time dependency (too annoying, can easily fit like any daft quant model) but I had a nonlinearity representing the response. Hence above comment, almost any model will do if I can get a perfect fit after half an hour doodling! Ditto kat's. 

They say 135 of the new deaths are due to a new diagnosis classification. Whatever that means. Were they all on one day? Are they going to reclassify old deaths as well? If so, do we just multiply all previous numbers by 2.5, say? 
 
User avatar
trackstar
Posts: 27504
Joined: August 28th, 2008, 1:53 pm

Re: How good or bad are the pandemic warning models?

February 13th, 2020, 4:44 pm


They say 135 of the new deaths are due to a new diagnosis classification. Whatever that means. Were they all on one day? Are they going to reclassify old deaths as well? If so, do we just multiply all previous numbers by 2.5, say? 
"What is the new diagnosis method?
The province now includes "clinically diagnosed cases" in the number of confirmed cases. 
This means it includes those showing symptoms, and having a CT scan showing an infected lung, rather than relying only on the standard nucleic acid tests.
Of the 242 new deaths in Wuhan, 135 are such "clinically diagnosed" cases. 
That means, even without the new definition, the number of deaths in Hubei on Wednesday was 107 - a new high for the province.
The province's 14,840 new infections include 13,332 clinically diagnosed cases. 
Overall, the province now has 48,206 confirmed infections."

Sharp increase in deaths and cases in Hubei - BBC News Feb 13

Also watching for medical papers.  This article (by your favorite news source!) is at least fairly detailed, balanced, and informative. 

Much of the reporting, especially since this new turn of events, has featured sensationalist and scary headlines and either not enough details or even wrong information - e.g. leaving out the CT scan, not explaining how such a change effects the stats etc.

**

""Clinically diagnosed cases" are those patients who demonstrate all the symptoms of Covid-19 but have been unable to be scientifically tested, or died before they were tested. The hope is that more people will be able to receive treatment by allowing doctors to diagnose them with the virus."

...big jump in cases as country expands diagnosis - CNN Feb 13

The WHO has issued special advisories for developing countries, as they may lack the resources and information flows to discern, address, and report any cases that emerge.

Scientists fear coronavirus spread in countries least able to contain it - Nature Feb 13

"The possibility of unreported cases is particularly concerning in countries with weaker health-care systems, such as those in southeast Asia and Africa, which could quickly be overwhelmed by a local outbreak, experts say. Although no cases have yet been reported in Africa, some countries there, such as Nigeria, are at particular risk because of their strong business ties to China.
Researchers have been using flight data to create models of the virus’s possible spread around the world. On 11 February, the International Committee on Taxonomy of Viruses officially named the virus SARS-CoV-2, highlighting its similarity to the virus that causes severe acute respiratory syndrome (SARS).
One model identified 30 countries or regions at risk of importing SARS-CoV-2 on the basis of the large number of flights from Wuhan, the outbreak’s epicentre, and from other cities in China with many travellers from Wuhan. The model used flight data from February 2018."

Also cited in that article from Nature: 

1) Lai, S. et al. Preprint on medRxiv https://www.medrxiv.org/content/10.1101 ... 20020479v1 (2020).
Assessing spread risk of Wuhan novel coronavirus within and beyond China, January-April 2020: a travel network-based modelling study
Shengjie Lai, Isaac Bogoch, Nick Ruktanonchai, Alexander Watts, Yu Li, Jianzing Yu, Xin Lv, Weizhong Yang, Hongjie Yu, Kamran Khan, Zhongjie Li, Andrew J Tatem

2) De Salazar, P. M., Niehus, R., Taylor, A., Buckee, C. & Lipsitch, M. Preprint on medRxiv https://doi.org/10.1101/2020.02.04.20020495 (2020).
Using predicted imports of 2019-nCoV cases to determine locations that may not be identifying all imported cases
Pablo M De Salazar, Rene Niehus, Aimee Taylor, Caroline O Buckee, Marc Lipsitch

3) Gilbert, M. et al. Preprint on medRxiv https://doi.org/10.1101/2020.02.05.20020792 (2020).
Preparedness and vulnerability of African countries against introductions of 2019-nCoV
Marius Gilbert, Giulia Pullano, Francesco Pinotti, Eugenio Valdano, Chiara Poletto, Pierre-Yves Boelle, Eric D'Ortenzio, Yazdan Yazdanpanah, Serge Paul Eholie, Mathias Altmann, Bernardo Gutierrez, Moritz U. G. Kraemer, Vittoria Colizza

Very long post and that's enough from me for now. But maybe useful for some of you who can critique models and build your own, as Paul is doing.
 
User avatar
trackstar
Posts: 27504
Joined: August 28th, 2008, 1:53 pm

Re: How good or bad are the pandemic warning models?

February 14th, 2020, 6:16 am

And then there is always human error...

They have now written off 108 of the reported deaths due to a problem with double counting.

https://www.cnbc.com/amp/2020/02/14/chi ... error.html

I’m going skiing - enjoy your weekends!
 
User avatar
Paul
Posts: 10796
Joined: July 20th, 2001, 3:28 pm

Re: How good or bad are the pandemic warning models?

February 14th, 2020, 6:55 am

The idiots.

So what were the number of deaths (using old method and new) for Feb 12 and 13? I am seeing lots of different numbers being thrown around!
 
User avatar
Paul
Posts: 10796
Joined: July 20th, 2001, 3:28 pm

Re: How good or bad are the pandemic warning models?

February 14th, 2020, 8:33 am

And the CDC test kits are "flawed"! These muppets make bankers look smart.

Thank goodness corona isn't that serious. (Am not sure why it's much different from flu in an old person turning into pneumonia.)
Every year an estimated 290,000 to 650,000 people die in the world due to complications from seasonal influenza (flu) viruses.

This figure corresponds to 795 to 1,781 deaths per day due to the seasonal flu.
 
User avatar
Cuchulainn
Posts: 62936
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: How good or bad are the pandemic warning models?

February 14th, 2020, 12:00 pm

And how many people die of 'old age' each year?
Step over the gap, not into it. Watch the space between platform and train.
http://www.datasimfinancial.com
http://www.datasim.nl
 
User avatar
tw
Posts: 1176
Joined: May 10th, 2002, 3:30 pm

Re: How good or bad are the pandemic warning models?

February 14th, 2020, 12:11 pm

And how many people die of 'old age' each year?
Life is a sexually transmitted disease & there’s a 100% mortality rate. —R. D. Laing
 
User avatar
katastrofa
Posts: 9588
Joined: August 16th, 2007, 5:36 am
Location: Alpha Centauri

Re: How good or bad are the pandemic warning models?

February 14th, 2020, 12:24 pm

Unless you're a jellyfish :-P

BTW, I always stop for a while at this: "human error" as distinguished from whose?
 
User avatar
Paul
Posts: 10796
Joined: July 20th, 2001, 3:28 pm

Re: How good or bad are the pandemic warning models?

February 14th, 2020, 10:25 pm

Fever, aches, shivers, dry cough. Bugger.
 
User avatar
trackstar
Posts: 27504
Joined: August 28th, 2008, 1:53 pm

Re: How good or bad are the pandemic warning models?

February 14th, 2020, 10:39 pm

Take care. I have been fighting off psychosomatic symptoms with a little pear brandy each night.

Has worked so far.  Also chicken soup and just staying warm under fresh blankets, even while working on stuff and posting here. 

Strength and courage to you! 
 
User avatar
Collector
Topic Author
Posts: 4734
Joined: August 21st, 2001, 12:37 pm

Re: How good or bad are the pandemic warning models?

February 14th, 2020, 10:59 pm

what bothers me is how certain western countries (bureaucrats) in particular, these days brag they have so good and modern health system so the corona will be no issue...

Fact check: mostly what they can do is isolation of patients, and they only have capacity for very few patients in high safe rooms etc. So NHS etc will be quickly overhelmed if a real break out of c-19, second they likely have even less access to safety equipment than china.

if get standard flue, hide under your bed and hope no one report u with fever and send u to hospital where one has considerably probability of getting infected by corona from the health care workers that are supposed to save ones life.

More than 1,700 health workers infected by coronavirus in China
"WHO is seeking information after country announced six health professionals had died"


remember also when many falls in the same week, it is potential increased risk of ending up as Lazarus statistic, after all, how much time can they waist spending on checking if people are really gone or just in self induced medical coma to heal. 
 
User avatar
trackstar
Posts: 27504
Joined: August 28th, 2008, 1:53 pm

Re: How good or bad are the pandemic warning models?

February 15th, 2020, 7:40 am

Taking a few days to let the numbers settle down, but looking at other corona news and this is interesting:

The WHO gave a press conference in Geneva yesterday and one of the the comments was this:

“Smoking’s Role in Coronavirus Merits Study, WHO Says (11:40 a.m. NY)

Smoking may play a role in the severity of infections caused by the coronavirus in China and merits more study, a World Health Organization official said Friday.

Chinese men have had more severe cases of the virus than women and also smoke at higher rates, Michael Ryan, executive director of the WHO Health Emergencies Programme, said at a press conference in Geneva.

The virus can cause pneumonia as it infects patients’ lungs, which can be weakened from cigarette use. China has higher overall rates of smoking than much of the rest of the world, according to WHO data.

“It’s an excellent hypothesis, but one that is unproven,” Ryan said. “There will be a lot of interest in looking at smoking as a risk factor.””

**

And on the numbers front, a paper published on Feb 7, so before the debacle this week, but still a useful summary, which contains four additional paper references (all 2020) at the end.

2019-Novel Coronavirus (2019-nCoV): estimating the case fatality rate – a word of caution
DOI: https://doi.org/10.4414/smw.2020.20203
Publication Date: 07.02.2020
Swiss Med Wkly. 2020;150:w20203

Battegay Manuela, Kuehl Richarda, Tschudin-Sutter Saraha, Hirsch Hans H.abc, Widmer Andreas F.a, Neher Richard A.d

https://smw.ch/article/doi/smw.2020.20203
 
User avatar
Cuchulainn
Posts: 62936
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: How good or bad are the pandemic warning models?

February 15th, 2020, 12:39 pm

Image
Step over the gap, not into it. Watch the space between platform and train.
http://www.datasimfinancial.com
http://www.datasim.nl
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