Bogleheads community discussion - Coronavirus

US Local Chapters, Wiki, and general Bogleheads community discussion, news, events, and announcements.
Locked
Unladen_Swallow
Posts: 784
Joined: Tue Dec 10, 2019 5:12 pm

Re: Bogleheads community discussion - Coronavirus

Post by Unladen_Swallow »

Preliminary findings from antibody testing in New York (source - Governors press presentation) reveals that NYC has a 21% infected rate. That is what, 3 million infected?


*multiple publications are estimating Fatality rate between 0.2-0.4. Not quite the 3-4% that.....
"I think it's much more interesting to live not knowing than to have answers which might be wrong." - Richard Feynman
JonnyB
Posts: 688
Joined: Sun Jan 19, 2020 4:28 pm

Re: Bogleheads community discussion - Coronavirus

Post by JonnyB »

Unladen_Swallow wrote: Thu Apr 23, 2020 4:07 pm Preliminary findings from antibody testing in New York (source - Governors press presentation) reveals that NYC has a 21% infected rate. That is what, 3 million infected?


*multiple publications are estimating Fatality rate between 0.2-0.4. Not quite the 3-4% that.....
Once again you need to reminded that Infection Fatality Rate and Case Fatality Rate are two different numbers.
dendylowen
Posts: 12
Joined: Fri Apr 20, 2018 5:07 pm

Re: Bogleheads community discussion - Coronavirus

Post by dendylowen »

Unladen_Swallow wrote: Thu Apr 23, 2020 4:07 pm Preliminary findings from antibody testing in New York (source - Governors press presentation) reveals that NYC has a 21% infected rate. That is what, 3 million infected?


*multiple publications are estimating Fatality rate between 0.2-0.4. Not quite the 3-4% that.....
A good news bad news situation. Good news- fatality rate is lower, bad news- R0 is much higher and more people will need to form immunity before herd immunity provides a benefit.
EnjoyIt
Posts: 8272
Joined: Sun Dec 29, 2013 7:06 pm

Re: Bogleheads community discussion - Coronavirus

Post by EnjoyIt »

ram wrote: Thu Apr 23, 2020 12:32 pm A person has a positive antibody test. Then what:
We do not know answers to many questions (at least I do not). Some reasonable guesses can be made for some questions.( Added: Answer to A can be found but need to read fine print)

Question A: Is the antibody test specific to SARS CoV2 or not
Option A1: Yes, it is specific to SARS CoV2
Option A2 : No, it tests antibodies against any corona virus.

Question B: Are antibodies protective against any strain of SARS CoV2 ?
Opt: B1: Yes . They are protective.
Opt B2: No. Not protective.

Question C: Are antibodies specific to one strain of SARS CoV protective for another strain of SARS CoV2
Option C1: Yes. Antibodies against one strain of Cov2 protect against all strains of SARS CoV2
Option C2: No antibodies against one strain are not protective for any other strain of SARS CoV 2
Option C3: Protective against some, but not all strains.

Question D: How long does the protectiveness (if any) last?
Option D1: Long lasting (months to years)
Option D2 : Short lasting (days to weeks)

Question E: Does T cell immunity develop and is protective against SARS CoV2 (T cell immunity is not measured by antibodies) https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4125530/
Option E1: Yes. It does and is protective.
Option E2: No. It does not.

There are many combinations for these options.

However the reassuring situation is A1, B1, C1, D1 and any E

To complicate matters further a negative antibody test with E1 is also reassuring.

My opinion: Antibody test may be modestly useful, but is unlikely to be the big answer to the timing of opening the economy. Other opinions welcome.
Thank you for laying it out so well. I fully agree that testing can offer some information, but I would not hang my hat on mass testing. The other potential benefit of testing is for serum donation. Once I find a place that does specific SARS-CoV2 antibody testing, I will get tested. The purpose is to donate serum if I am positive. But, I will still continue to wear PPE in the hospital. We just don't have enough information to change my actions.
A time to EVALUATE your jitters: | viewtopic.php?p=1139732#p1139732
EnjoyIt
Posts: 8272
Joined: Sun Dec 29, 2013 7:06 pm

Re: Bogleheads community discussion - Coronavirus

Post by EnjoyIt »

JonnyB wrote: Thu Apr 23, 2020 12:52 pm
AlphaLess wrote: Thu Apr 23, 2020 11:45 am
A. highly infected place like NYC, which according to Cuomo has a 20% infection rate (and according to the model I posted, is likely to be true),
B. highly sparsely population place with strict rules on some activities (like eating at a restaurant and getting a haircut).

Those in the middle are stuck.

NYC can open up because they have already paid the price.
So assuming a 20% infection rate, what about the other 80%? Do you just write them off? Most models say you need around 80% for herd immunity. So NYC isn't even close. You presumably need four times as many infections and four times as many deaths to get there. Where does your conclusion that NYC can open up come from?

And outside of the boroughs of NYC, the same study showed an infection rate of only 3%. How does opening NYC work with hundreds of thousands moving back and forth from NYC to the rest of the state every day?

The second wave of the 1918 Spanish flu was much worst than the first. It takes a lot of infections and deaths to get to herd immunity.
I would hope the remaining 80% has acquired masks (to decrease risk of inhalation of high viral loads) and practices good hand hygiene so that they can protect themselves. Also, if 20% of the population is immune (and that is a big IF) then the R0 should be lower going forward. That in addition to a slow opening up of our economy, I would suspect that patients will continue to get sick but at a rate that will not overwhelm the healthcare system. It will be imperative that people protect themselves.

Just being out in public around COVID patients will not necessarily get one sick if they have the ability to decrease their viral load inoculation. Plenty of healthcare providers are not getting sick and they are definitely in contact with COVID patients. Many healthcare providers outside of hotspots and COVID wards are just wearing surgical mask except in high risk situations and not getting sick. Understand also that there have been a good amount of healthcare providers who have gotten sick which is why again, it is so important to take precautions.
Last edited by EnjoyIt on Thu Apr 23, 2020 4:57 pm, edited 1 time in total.
A time to EVALUATE your jitters: | viewtopic.php?p=1139732#p1139732
mouses
Posts: 4217
Joined: Sat Oct 24, 2015 12:24 am

Re: Bogleheads community discussion - Coronavirus

Post by mouses »

folkher0 wrote: Thu Apr 23, 2020 10:27 am
The last thing I will say about this thing that is kind of weird is that I don't know anything about who made this model. Here's my academic bias again but usually I would expect someone to attach their credentials to a model. All I see is a name Youyang Gu. It is attached to a Twitter feed that has a picture of a guy and says data scientist MIT '15.

That's fine. I have no reason to believe or not believe that. I don't think a project like this is deceptive or fake, just that usually modeling projects like this involve a team of people collaborating, discussing, tweaking, etc. One guy with a computer is....fine. Probably not what I would hang my hat on in terms of policy, but...fine.
Well, such a person exists, but I don't know if he built this model. From the alumni directory:

New York, New York
Mr. Youyang Gu '15, MNG '16
Trader
2015, SB - Bachelor Of Science, 18 - Mathematics
2015, SB - Bachelor Of Science, 62 - Elec Eng & Comp Sci
2016, MNG - Master of Engineering, 6P - Elec Eng & Comp Sci

I have no objection to one person projects.
EnjoyIt
Posts: 8272
Joined: Sun Dec 29, 2013 7:06 pm

Re: Bogleheads community discussion - Coronavirus

Post by EnjoyIt »

folkher0 wrote: Thu Apr 23, 2020 2:08 pm
Teague wrote: Thu Apr 23, 2020 2:04 pm ^ Cuomo noted the uncertainties involved and admits he does not know what to make of this data at this time.

(The sample was definitely not random, they sampled supermarket patrons in some fashion.)
Yeah but better than I might have thought. It’s not totally random but really not bad all things considered.
That study therefor excludes people sick in hospitals or those who have died. Don't know how that would affect the data.

it also excludes those who refuse to leave home and have everything delivered.
A time to EVALUATE your jitters: | viewtopic.php?p=1139732#p1139732
User avatar
wshang
Posts: 1239
Joined: Sun Feb 03, 2008 10:40 am

Re: Bogleheads community discussion - Coronavirus

Post by wshang »

EnjoyIt wrote: Thu Apr 23, 2020 4:43 pm I would hope the remaining 80% has acquired masks (to decrease risk of inhalation of high viral loads)
You are not the only colleague who has been repeating the mantra "high viral load". I have been searching in vain for studies other than analogies to known diseases in which this occurs, (influenza and HIV), BUT have not been able to find supporting evidence.

Admittedly it is a seductive idea which dovetails to the more established theory of ACE2 receptor upregulation correlating with disease severity. Can you share any studies other than medical or popular press speculative OpEd's?
EnjoyIt
Posts: 8272
Joined: Sun Dec 29, 2013 7:06 pm

Re: Bogleheads community discussion - Coronavirus

Post by EnjoyIt »

wshang wrote: Thu Apr 23, 2020 4:57 pm
EnjoyIt wrote: Thu Apr 23, 2020 4:43 pm I would hope the remaining 80% has acquired masks (to decrease risk of inhalation of high viral loads)
You are not the only colleague who has been repeating the mantra "high viral load". I have been searching in vain for studies other than analogies to known diseases in which this occurs, (influenza and HIV), BUT have not been able to find supporting evidence.

Admittedly it is a seductive idea which dovetails to the more established theory of ACE2 receptor upregulation correlating with disease severity. Can you share any studies other than medical or popular press speculative OpEd's?
That's the problem. How do you do a study like this? Give cohort A 10,000 viruses, Cohort B 100,000 viruses, C 1,000,000 viruses, D control? That just doesn't sound ethical to me.

I honestly infer this information from what we are seeing. If viral load did not matter, I and many of my colleagues and nurses would have surely been sick by now (unless we are all some of those asymptomatic people if viral load did not matter.) Most of use wear surgical masks unless we are going into a high risk room. Many of us have come in contact with patients later to be diagnosed with COVID only wearing a surgical mask and we have not gotten sick. As we all know, surgical masks are not as good as N95 masks that are properly worn.

I also use viral load in my discussion above because surgical masks are not a 100% barrier and I do not want people reading this to infer that it is.
A time to EVALUATE your jitters: | viewtopic.php?p=1139732#p1139732
AlphaLess
Posts: 3409
Joined: Fri Sep 29, 2017 11:38 pm
Location: Kentucky

Re: Bogleheads community discussion - Coronavirus

Post by AlphaLess »

JonnyB wrote: Thu Apr 23, 2020 12:52 pm
AlphaLess wrote: Thu Apr 23, 2020 11:45 am
A. highly infected place like NYC, which according to Cuomo has a 20% infection rate (and according to the model I posted, is likely to be true),
B. highly sparsely population place with strict rules on some activities (like eating at a restaurant and getting a haircut).

Those in the middle are stuck.

NYC can open up because they have already paid the price.
So assuming a 20% infection rate, what about the other 80%? Do you just write them off? Most models say you need around 80% for herd immunity. So NYC isn't even close. You presumably need four times as many infections and four times as many deaths to get there. Where does your conclusion that NYC can open up come from?

And outside of the boroughs of NYC, the same study showed an infection rate of only 3%. How does opening NYC work with hundreds of thousands moving back and forth from NYC to the rest of the state every day?

The second wave of the 1918 Spanish flu was much worst than the first. It takes a lot of infections and deaths to get to herd immunity.
So, herd immunity and R0 are intimately related.

I think there were reports that the R0 of this thing in large, integrated cities with public transit is around 5.7. So, that translates to herd immunity requirement of 83%. At the same time, with some distancing measures, R0 can be reduced, so, say, 2.5. At that point, herd immunity requirement is 60%.

The key is NOT to overwhelm the health care system, because in that case, fatality rate can go higher.

There are estimates that with the strict social distancing measures, R0 in NYC is around 0.8 currently. (1-R0^n) / ( 1 - R0), where N is the number of epochs, and an epoch is around 5-6 days.

Currently infecteds are 2.3% in NY, and with an R0 of 0.8, this thing will stop when another 10% or so get infected, which will take a long number of epochs.

There is no way to prevent a second wave from happening, unless there is either herd immunity, or a properly functioning vaccine.
I don't carry a signature because people are easily offended.
AlphaLess
Posts: 3409
Joined: Fri Sep 29, 2017 11:38 pm
Location: Kentucky

Re: Bogleheads community discussion - Coronavirus

Post by AlphaLess »

webbie90 wrote: Thu Apr 23, 2020 12:07 pm
AlphaLess wrote: Thu Apr 23, 2020 11:45 am
MtnBiker wrote: Thu Apr 23, 2020 11:23 am
AlphaLess wrote: Wed Apr 22, 2020 9:35 pm Another COVID projection model. I tend to like this one a lot more.

#1. It is non-parametric, i.e., does not make any assumptions about the type of curve,
#2. Does not make assumptions about R0. Rather, infers R0,
#3. Models hidden variables such as true number infected,
#4. Updates in real-time.

https://covid19-projections.com/

https://covid19-projections.com/about/
A major difference with this model and the IMHE model seems to be that the latter fails to account for the continuing (lower) rate of infections after the first wave. I think the results displayed now for both assume fairly strict social distancing will continue through the end of May.

As we know, some states are going to start loosening social distancing standards much sooner than that. It will be interesting to see what the models predict will happen in states which "open up" sooner rather than later.
There is room to open up on two sides of the spectrum:

A. highly infected place like NYC, which according to Cuomo has a 20% infection rate (and according to the model I posted, is likely to be true),
B. highly sparsely population place with strict rules on some activities (like eating at a restaurant and getting a haircut).

Those in the middle are stuck.

NYC can open up because they have already paid the price.

At any rate, without injecting my own opinion about opening up or not, I applaud those states that are brave enough to try it. Of course, they need to live with the consequences of their actions too.
If Cuomo is right and 20% of NYC residents have been infected, that would mean about 1.7 million infected (20% x 8.4 million residents of NYC). As of yesterday there were 15,000 deaths in NYC if you count probable cases who had symptoms but were never tested. That works out to an infection fatality ratio of 0.9%. Of course a lot hinges on the size of the denominator, but an infection fatality ratio of around 1% is not far off from estimates in countries that have done a lot of testing per capita.
I agree with this estimates. I have long thought that the IFR of this thing is 0.85% - 1.00%.

I also agree with Cuomo w/r to the infection table he put up today.

PCR-positives in NYS are 235K. So 2.5M infecteds is NOT that hard to believe, considering the high positive/tested ratio.

Also, see this model, which infers (predicts) the true rate of infections:

https://covid19-projections.com/us-ny

Total infected as of April 24th in NYS: 12.9% which is right in line with Cuomo.
I don't carry a signature because people are easily offended.
AlphaLess
Posts: 3409
Joined: Fri Sep 29, 2017 11:38 pm
Location: Kentucky

Re: Bogleheads community discussion - Coronavirus

Post by AlphaLess »

folkher0 wrote: Thu Apr 23, 2020 12:01 pm The model is “updated in real time”. It ought to be close. Today.
Exactly. Which is a HUGE value added.

Prior to Cuomo's revelation of the serum-test driven numbers TODAY, the model had predictions of the hidden number of infecteds.

And that number is very close to what Cuomo put up today (I think it is slightly higher, but we are arguing about trees in a forest).
I don't carry a signature because people are easily offended.
Unladen_Swallow
Posts: 784
Joined: Tue Dec 10, 2019 5:12 pm

Re: Bogleheads community discussion - Coronavirus

Post by Unladen_Swallow »

dendylowen wrote: Thu Apr 23, 2020 4:18 pm
Unladen_Swallow wrote: Thu Apr 23, 2020 4:07 pm Preliminary findings from antibody testing in New York (source - Governors press presentation) reveals that NYC has a 21% infected rate. That is what, 3 million infected?


*multiple publications are estimating Fatality rate between 0.2-0.4. Not quite the 3-4% that.....
A good news bad news situation. Good news- fatality rate is lower, bad news- R0 is much higher and more people will need to form immunity before herd immunity provides a benefit.
Yes. Objectively it is good news/bad news. But compared to the reaction so far, I find it good news. I think the reaction has served the hysteria, but has done a poor job serving the vulnerable.

Another unrelated point - many compare cases in different countries and their performance. I don't think we can truly understand this until a few months from now. No two countries are identical. Demographics, weather, natural immunity, medications etc could play a significant role. Taking positive cues or negative cues are preliminary. Some countries that do well will do so in spite of their actions (or lack thereof), not because of it.
"I think it's much more interesting to live not knowing than to have answers which might be wrong." - Richard Feynman
Unladen_Swallow
Posts: 784
Joined: Tue Dec 10, 2019 5:12 pm

Re: Bogleheads community discussion - Coronavirus

Post by Unladen_Swallow »

JonnyB wrote: Thu Apr 23, 2020 4:13 pm
Unladen_Swallow wrote: Thu Apr 23, 2020 4:07 pm Preliminary findings from antibody testing in New York (source - Governors press presentation) reveals that NYC has a 21% infected rate. That is what, 3 million infected?


*multiple publications are estimating Fatality rate between 0.2-0.4. Not quite the 3-4% that.....
Once again you need to reminded that Infection Fatality Rate and Case Fatality Rate are two different numbers.
If you are confused, I can help.
"I think it's much more interesting to live not knowing than to have answers which might be wrong." - Richard Feynman
User avatar
wshang
Posts: 1239
Joined: Sun Feb 03, 2008 10:40 am

Re: Bogleheads community discussion - Coronavirus

Post by wshang »

EnjoyIt wrote: Thu Apr 23, 2020 5:05 pm
wshang wrote: Thu Apr 23, 2020 4:57 pm
EnjoyIt wrote: Thu Apr 23, 2020 4:43 pm I would hope the remaining 80% has acquired masks (to decrease risk of inhalation of high viral loads)
You are not the only colleague who has been repeating the mantra "high viral load".
If viral load did not matter, I and many of my colleagues and nurses would have surely been sick by now (unless we are all some of those asymptomatic people if viral load did not matter.)
There is a pathophysiologic rationale for dose dependency. Droplets get trapped for the most part in the nasopharynx. The turbinates create turbulence to send about 80% to the sticky mucus. If that's true, then very few of the ~100 nanometer sized virions make it to the lungs.

It does seem for many that the infection is two-tiered. Beat the infection in the nose, avoid the lung phase and you are golden. That would also provide a rationale for the dichotomy we are seeing between the warmer temperature infection rates and severe disease. T-cell mucosal immunity rules up top, false negative test results?. Finally, if you get a large viral dose to the lung, the ACE2 receptors get widely saturated, widespread ACE2 downregulation then coincides with ARDS occurs. The last part admittedly gets really speculative though.
folkher0
Posts: 912
Joined: Fri Dec 09, 2016 1:48 pm

Re: Bogleheads community discussion - Coronavirus

Post by folkher0 »

AlphaLess wrote: Thu Apr 23, 2020 5:15 pm
folkher0 wrote: Thu Apr 23, 2020 12:01 pm The model is “updated in real time”. It ought to be close. Today.
Exactly. Which is a HUGE value added.

Prior to Cuomo's revelation of the serum-test driven numbers TODAY, the model had predictions of the hidden number of infecteds.

And that number is very close to what Cuomo put up today (I think it is slightly higher, but we are arguing about trees in a forest).
I gotta be honest it’s a really slick looking site. But it’s just weird. The guy isn’t really transparent about the covariates he’s using, how the model was constructed outside generally referencing some methodology, and uses buzzy terms like “machine learning” (I don’t really know what that means) and “real-time” updates.

Models are usually validated. Both internally (using a subset of the data in the model to make an inference about the rest of the data included) and externally (applying the model to outside data to see if associations hold up, usually retrospectively). I didn’t see that on the page, but I could be wrong and just missed it. All he says is “we have a strong Validation blah blah”

There’s a bunch of other things like that.

And it’s dot com site.

I guess what I’m getting at is that I have no idea if it’s a good model or not, owever accurate it may look at any one time. To use a bogleheads reference it’s like data mining in portfolio visualizer to find the magic combination of stocks and funds in retrospect to show what would have made you a billionaire in 1980 if only you could buy small cap value...

Anyway. It might be great. Might be better than any other model. I don’t know cus I can’t really evaluate it.

My vibe is that the dude is marketing himself for a job. That’s fine. Maybe he’s a superstar. Maybe this is THE answer. I don’t know because he didn’t show us his work.
veggivet
Posts: 1480
Joined: Sun Jan 25, 2015 12:07 pm
Location: New England

Re: Bogleheads community discussion - Coronavirus

Post by veggivet »

wshang wrote: Thu Apr 23, 2020 5:27 pm
EnjoyIt wrote: Thu Apr 23, 2020 5:05 pm
wshang wrote: Thu Apr 23, 2020 4:57 pm
EnjoyIt wrote: Thu Apr 23, 2020 4:43 pm I would hope the remaining 80% has acquired masks (to decrease risk of inhalation of high viral loads)
You are not the only colleague who has been repeating the mantra "high viral load".
If viral load did not matter, I and many of my colleagues and nurses would have surely been sick by now (unless we are all some of those asymptomatic people if viral load did not matter.)
There is a pathophysiologic rationale for dose dependency. Droplets get trapped for the most part in the nasopharynx. The turbinates create turbulence to send about 80% to the sticky mucus. If that's true, then very few of the ~100 nanometer sized virions make it to the lungs.

It does seem for many that the infection is two-tiered. Beat the infection in the nose, avoid the lung phase and you are golden. That would also provide a rationale for the dichotomy we are seeing between the warmer temperature infection rates and severe disease. T-cell mucosal immunity rules up top, false negative test results?. Finally, if you get a large viral dose to the lung, the ACE2 receptors get widely saturated, widespread ACE2 downregulation then coincides with ARDS occurs. The last part admittedly gets really speculative though.
Taking this one step further, would a nasal vaccine that stimulates IgA production be superior, at least in concept, to a parenteral vaccine? I believe the influenza vaccines help protect against infection at least in part due to nasal IgA production.
User avatar
baconavocado
Posts: 787
Joined: Fri Oct 13, 2017 3:03 pm

Re: Bogleheads community discussion - Coronavirus

Post by baconavocado »

Hopes dashed as coronavirus drug remdesivir fails first trial

https://www.bbc.com/news/world-52406261
folkher0
Posts: 912
Joined: Fri Dec 09, 2016 1:48 pm

Re: Bogleheads community discussion - Coronavirus

Post by folkher0 »

baconavocado wrote: Thu Apr 23, 2020 6:20 pm Hopes dashed as coronavirus drug remdesivir fails first trial

https://www.bbc.com/news/world-52406261
So much speculation on this drug based on one leaked conference call from Chicago and now “accidentally” published data from China.

I’ll reserve judgement either way for now.
EnjoyIt
Posts: 8272
Joined: Sun Dec 29, 2013 7:06 pm

Re: Bogleheads community discussion - Coronavirus

Post by EnjoyIt »

wshang wrote: Thu Apr 23, 2020 5:27 pm
EnjoyIt wrote: Thu Apr 23, 2020 5:05 pm
wshang wrote: Thu Apr 23, 2020 4:57 pm
EnjoyIt wrote: Thu Apr 23, 2020 4:43 pm I would hope the remaining 80% has acquired masks (to decrease risk of inhalation of high viral loads)
You are not the only colleague who has been repeating the mantra "high viral load".
If viral load did not matter, I and many of my colleagues and nurses would have surely been sick by now (unless we are all some of those asymptomatic people if viral load did not matter.)
There is a pathophysiologic rationale for dose dependency. Droplets get trapped for the most part in the nasopharynx. The turbinates create turbulence to send about 80% to the sticky mucus. If that's true, then very few of the ~100 nanometer sized virions make it to the lungs.

It does seem for many that the infection is two-tiered. Beat the infection in the nose, avoid the lung phase and you are golden. That would also provide a rationale for the dichotomy we are seeing between the warmer temperature infection rates and severe disease. T-cell mucosal immunity rules up top, false negative test results?. Finally, if you get a large viral dose to the lung, the ACE2 receptors get widely saturated, widespread ACE2 downregulation then coincides with ARDS occurs. The last part admittedly gets really speculative though.
So many ifs. I am no microbiologist/virologist but doesn't SARs-CoV2 viral particles enter the blood stream as well? Isn't that is why we are seeing GI symptoms (nausea, vomiting, and diarrhea,) pulmonary symptoms (cough and shortness of breath,) myocarditis, hematologic with low lymphocytes and platelets.

It looks like the second tier you describe may be viremic (viruses in the blood stream) than just settling down into the lungs.

I would assume if we are able to suppress viral expression/replication quickly by our primary defense in our respiratory mucous membranes such as our cilia, the mucous, and if my memory serves me right, IgA. If we can mount a solid defense there then we don't get viremia and don't end up in the ICU. I think that's why viral load plays a big role and why everyone wearing surgical masks makes a big difference.
A time to EVALUATE your jitters: | viewtopic.php?p=1139732#p1139732
User avatar
ram
Posts: 2281
Joined: Tue Jan 01, 2008 9:47 pm
Location: Midwest

Re: Bogleheads community discussion - Coronavirus

Post by ram »

EnjoyIt wrote: Thu Apr 23, 2020 5:05 pm
wshang wrote: Thu Apr 23, 2020 4:57 pm
EnjoyIt wrote: Thu Apr 23, 2020 4:43 pm I would hope the remaining 80% has acquired masks (to decrease risk of inhalation of high viral loads)
You are not the only colleague who has been repeating the mantra "high viral load". I have been searching in vain for studies other than analogies to known diseases in which this occurs, (influenza and HIV), BUT have not been able to find supporting evidence.

Admittedly it is a seductive idea which dovetails to the more established theory of ACE2 receptor upregulation correlating with disease severity. Can you share any studies other than medical or popular press speculative OpEd's?
That's the problem. How do you do a study like this? Give cohort A 10,000 viruses, Cohort B 100,000 viruses, C 1,000,000 viruses, D control? That just doesn't sound ethical to me.
I believe tracheostomies (holes in wind pipe) were made in monkeys and non human primates to study effects of smoking. I assume that lab animals would need to be the starting point of any such experiments.

John Hunter (English surgeon) allegedly used himself as a test subject for medical tests that were ethically difficult. https://en.wikipedia.org/wiki/John_Hunter_(surgeon)

Perhaps use something along the lines of radiation monitoring. Put a Virus absorbing/ adsorbing strip on healthcare workers and at the end of the shift measure virus exposure. Lets say the exposure turns out to Intensivist (1000,000) , Hospitalist (100,000), Outpatient clinic doctor (10,000) and now (roughly) you got your groups. Or you can take 4 groups by quartile of the exposure and proceed from there. I do not know what one can use to get the viral exposure counts. Maybe use viral counts on discarded masks. Just thinking out loud.
Ram
AlphaLess
Posts: 3409
Joined: Fri Sep 29, 2017 11:38 pm
Location: Kentucky

Re: Bogleheads community discussion - Coronavirus

Post by AlphaLess »

I just wanted to chime in and say that I am enjoying the NFL Draft, coronavirus edition.
I don't carry a signature because people are easily offended.
AlphaLess
Posts: 3409
Joined: Fri Sep 29, 2017 11:38 pm
Location: Kentucky

Re: Bogleheads community discussion - Coronavirus

Post by AlphaLess »

folkher0 wrote: Thu Apr 23, 2020 5:33 pm
AlphaLess wrote: Thu Apr 23, 2020 5:15 pm
folkher0 wrote: Thu Apr 23, 2020 12:01 pm The model is “updated in real time”. It ought to be close. Today.
Exactly. Which is a HUGE value added.

Prior to Cuomo's revelation of the serum-test driven numbers TODAY, the model had predictions of the hidden number of infecteds.

And that number is very close to what Cuomo put up today (I think it is slightly higher, but we are arguing about trees in a forest).
I gotta be honest it’s a really slick looking site. But it’s just weird. The guy isn’t really transparent about the covariates he’s using, how the model was constructed outside generally referencing some methodology, and uses buzzy terms like “machine learning” (I don’t really know what that means) and “real-time” updates.

Models are usually validated. Both internally (using a subset of the data in the model to make an inference about the rest of the data included) and externally (applying the model to outside data to see if associations hold up, usually retrospectively). I didn’t see that on the page, but I could be wrong and just missed it. All he says is “we have a strong Validation blah blah”

There’s a bunch of other things like that.

And it’s dot com site.

I guess what I’m getting at is that I have no idea if it’s a good model or not, owever accurate it may look at any one time. To use a bogleheads reference it’s like data mining in portfolio visualizer to find the magic combination of stocks and funds in retrospect to show what would have made you a billionaire in 1980 if only you could buy small cap value...

Anyway. It might be great. Might be better than any other model. I don’t know cus I can’t really evaluate it.

My vibe is that the dude is marketing himself for a job. That’s fine. Maybe he’s a superstar. Maybe this is THE answer. I don’t know because he didn’t show us his work.
I can certainly see your concerns. We don't know if it is a good model or not. But the general design is a heck-of-a-lot better than the IHME. You need to make assumptions to build a model, and some assumptions matter more than others. If you get the key assumptions wrong, you end up with something that is hard to believe.

Based on the information I have read on the website, I can answer some of your questions:
- the driving model is based on the SEIS model, https://en.wikipedia.org/wiki/Compartme ... SEIS_model ,
- "machine learning": that just means intermediate statistics,
- it appears that the key covariate is death count,
- real-time means that if / when there is new data, it take 10 minutes to generate the entire thing (as opposed to IHME, which can only be done 3 times a week.),

Quoting:

"The parameters/inputs of this simulator are then learned using machine learning techniques that attempts to minimize the error between the projected outputs and the actual results. We utilize daily deaths data reported by each state/country to forecast future deaths. After some additional validation techniques (to minimize a phenomenon called overfitting), we use the learned parameters to simulate the future and make projections."

Biggest criticism this author has with respect to IHME model is the Gaussian error assumption. This results in unrealistic, symmetric curves in the up and down stages.
I don't carry a signature because people are easily offended.
AlphaLess
Posts: 3409
Joined: Fri Sep 29, 2017 11:38 pm
Location: Kentucky

Re: Bogleheads community discussion - Coronavirus

Post by AlphaLess »

ram wrote: Thu Apr 23, 2020 6:37 pm Perhaps use something along the lines of radiation monitoring. Put a Virus absorbing/ adsorbing strip on healthcare workers and at the end of the shift measure virus exposure. Lets say the exposure turns out to Intensivist (1000,000) , Hospitalist (100,000), Outpatient clinic doctor (10,000) and now (roughly) you got your groups. Or you can take 4 groups by quartile of the exposure and proceed from there. I do not know what one can use to get the viral exposure counts. Maybe use viral counts on discarded masks. Just thinking out loud.
These are good points. Freakonomics has a lot of examples where naturally, control and test groups are formed.
I don't carry a signature because people are easily offended.
AlphaLess
Posts: 3409
Joined: Fri Sep 29, 2017 11:38 pm
Location: Kentucky

Re: Bogleheads community discussion - Coronavirus

Post by AlphaLess »

folkher0 wrote: Thu Apr 23, 2020 6:25 pm
baconavocado wrote: Thu Apr 23, 2020 6:20 pm Hopes dashed as coronavirus drug remdesivir fails first trial

https://www.bbc.com/news/world-52406261
So much speculation on this drug based on one leaked conference call from Chicago and now “accidentally” published data from China.

I’ll reserve judgement either way for now.
I think it is a mistake to look for a silver bullet.

Take Tamiflu, and how and how well it works.
I don't carry a signature because people are easily offended.
AlphaLess
Posts: 3409
Joined: Fri Sep 29, 2017 11:38 pm
Location: Kentucky

Re: Bogleheads community discussion - Coronavirus

Post by AlphaLess »

folkher0 wrote: Thu Apr 23, 2020 10:27 am
The last thing I will say about this thing that is kind of weird is that I don't know anything about who made this model. Here's my academic bias again but usually I would expect someone to attach their credentials to a model. All I see is a name Youyang Gu. It is attached to a Twitter feed that has a picture of a guy and says data scientist MIT '15.

That's fine. I have no reason to believe or not believe that. I don't think a project like this is deceptive or fake, just that usually modeling projects like this involve a team of people collaborating, discussing, tweaking, etc. One guy with a computer is....fine. Probably not what I would hang my hat on in terms of policy, but...fine.

Contact
We encourage questions/insights/feedback! Please reach out to Youyang Gu on Twitter via @youyanggu or on LinkedIn.
I don't carry a signature because people are easily offended.
folkher0
Posts: 912
Joined: Fri Dec 09, 2016 1:48 pm

Re: Bogleheads community discussion - Coronavirus

Post by folkher0 »

AlphaLess wrote: Thu Apr 23, 2020 7:07 pm I just wanted to chime in and say that I am enjoying the NFL Draft, coronavirus edition.
Our last conversation reminded me. I took some time off in residency and got a masters degree. Lots of biostats. So I downloaded like 20 years of baseball stats and built this crazy model yada yada yada. My fantasy draft was auto draft based on the model. I won my fantasy league that year. I thought about quitting surgery to be the next Billy Beane.

I ran the model the next year and came in dead last.

Turned out the year I won I had just got lucky.

Glad I stuck with Surgery.
Last edited by folkher0 on Thu Apr 23, 2020 7:40 pm, edited 1 time in total.
User avatar
ResearchMed
Posts: 16795
Joined: Fri Dec 26, 2008 10:25 pm

Re: Bogleheads community discussion - Coronavirus

Post by ResearchMed »

AlphaLess wrote: Thu Apr 23, 2020 7:20 pm
folkher0 wrote: Thu Apr 23, 2020 6:25 pm
baconavocado wrote: Thu Apr 23, 2020 6:20 pm Hopes dashed as coronavirus drug remdesivir fails first trial

https://www.bbc.com/news/world-52406261
So much speculation on this drug based on one leaked conference call from Chicago and now “accidentally” published data from China.

I’ll reserve judgement either way for now.
I think it is a mistake to look for a silver bullet.

Take Tamiflu, and how and how well it works.
Would Tamiflu work on the novel coronavirus?
If so, why haven't we heard wondrous tales about the successes?

RM
This signature is a placebo. You are in the control group.
folkher0
Posts: 912
Joined: Fri Dec 09, 2016 1:48 pm

Re: Bogleheads community discussion - Coronavirus

Post by folkher0 »

AlphaLess wrote: Thu Apr 23, 2020 7:17 pm
folkher0 wrote: Thu Apr 23, 2020 5:33 pm
AlphaLess wrote: Thu Apr 23, 2020 5:15 pm
folkher0 wrote: Thu Apr 23, 2020 12:01 pm The model is “updated in real time”. It ought to be close. Today.
Exactly. Which is a HUGE value added.

Prior to Cuomo's revelation of the serum-test driven numbers TODAY, the model had predictions of the hidden number of infecteds.

And that number is very close to what Cuomo put up today (I think it is slightly higher, but we are arguing about trees in a forest).
I gotta be honest it’s a really slick looking site. But it’s just weird. The guy isn’t really transparent about the covariates he’s using, how the model was constructed outside generally referencing some methodology, and uses buzzy terms like “machine learning” (I don’t really know what that means) and “real-time” updates.

Models are usually validated. Both internally (using a subset of the data in the model to make an inference about the rest of the data included) and externally (applying the model to outside data to see if associations hold up, usually retrospectively). I didn’t see that on the page, but I could be wrong and just missed it. All he says is “we have a strong Validation blah blah”

There’s a bunch of other things like that.

And it’s dot com site.

I guess what I’m getting at is that I have no idea if it’s a good model or not, owever accurate it may look at any one time. To use a bogleheads reference it’s like data mining in portfolio visualizer to find the magic combination of stocks and funds in retrospect to show what would have made you a billionaire in 1980 if only you could buy small cap value...

Anyway. It might be great. Might be better than any other model. I don’t know cus I can’t really evaluate it.

My vibe is that the dude is marketing himself for a job. That’s fine. Maybe he’s a superstar. Maybe this is THE answer. I don’t know because he didn’t show us his work.
I can certainly see your concerns. We don't know if it is a good model or not. But the general design is a heck-of-a-lot better than the IHME. You need to make assumptions to build a model, and some assumptions matter more than others. If you get the key assumptions wrong, you end up with something that is hard to believe.

Based on the information I have read on the website, I can answer some of your questions:
- the driving model is based on the SEIS model, https://en.wikipedia.org/wiki/Compartme ... SEIS_model ,
- "machine learning": that just means intermediate statistics,
- it appears that the key covariate is death count,
- real-time means that if / when there is new data, it take 10 minutes to generate the entire thing (as opposed to IHME, which can only be done 3 times a week.),

Quoting:

"The parameters/inputs of this simulator are then learned using machine learning techniques that attempts to minimize the error between the projected outputs and the actual results. We utilize daily deaths data reported by each state/country to forecast future deaths. After some additional validation techniques (to minimize a phenomenon called overfitting), we use the learned parameters to simulate the future and make projections."

Biggest criticism this author has with respect to IHME model is the Gaussian error assumption. This results in unrealistic, symmetric curves in the up and down stages.
Right nonparametric is fine. I’m not even saying it’s a bad model. I’m just saying I don’t know what he did. I can read the words. But there are no numbers.

I’ll give you an example. In surgery there are a thousand databases for each field. Trauma thoracic transplant cancer etc. Each of those databases generate models to correlate outcomes to various predictors (hospital size, volume, patient comorbitities etc). You can go to those websites, look at the models and see a list of the variables that were included, why the variables were included and the regression coefficients associated with those covariates.

That way you can, if you want, build the model yourself.

Every paper you read which models an outcome (death) for instance) does the same thing. That’s what I mean by showing your work.you even have to report the type and version of the stats software that you use. If no one can recreate what what you’ve done, it doesn’t count.

Harsh I know. But that’s the game.
folkher0
Posts: 912
Joined: Fri Dec 09, 2016 1:48 pm

Re: Bogleheads community discussion - Coronavirus

Post by folkher0 »

ResearchMed wrote: Thu Apr 23, 2020 7:23 pm
AlphaLess wrote: Thu Apr 23, 2020 7:20 pm
folkher0 wrote: Thu Apr 23, 2020 6:25 pm
baconavocado wrote: Thu Apr 23, 2020 6:20 pm Hopes dashed as coronavirus drug remdesivir fails first trial

https://www.bbc.com/news/world-52406261
So much speculation on this drug based on one leaked conference call from Chicago and now “accidentally” published data from China.

I’ll reserve judgement either way for now.
I think it is a mistake to look for a silver bullet.

Take Tamiflu, and how and how well it works.
Would Tamiflu work on the novel coronavirus?
If so, why haven't we heard wondrous tales about the successes?

RM
No. Tamiflu doesn’t work. It doesn’t even do much for the flu. I think he/she is being sarcastic.
AlphaLess
Posts: 3409
Joined: Fri Sep 29, 2017 11:38 pm
Location: Kentucky

Re: Bogleheads community discussion - Coronavirus

Post by AlphaLess »

ResearchMed wrote: Thu Apr 23, 2020 7:23 pm
AlphaLess wrote: Thu Apr 23, 2020 7:20 pm
folkher0 wrote: Thu Apr 23, 2020 6:25 pm
baconavocado wrote: Thu Apr 23, 2020 6:20 pm Hopes dashed as coronavirus drug remdesivir fails first trial

https://www.bbc.com/news/world-52406261
So much speculation on this drug based on one leaked conference call from Chicago and now “accidentally” published data from China.

I’ll reserve judgement either way for now.
I think it is a mistake to look for a silver bullet.

Take Tamiflu, and how and how well it works.
Would Tamiflu work on the novel coronavirus?
If so, why haven't we heard wondrous tales about the successes?

RM
Sorry, I should have been more clear. When bringing in Tamiflu, I was referring with existing body of research of that drug with influenza.

My points are the following:
- Tamiflu works if administered early (within 48 hours of symptoms),
- Tamiflu reduces severity and length of symptoms in non-critical cases,
- Tamiflu does not affect hospitalization rates.

So, from the above one could say that Tamiflu is useless or that Tamiflu is useful, depending on what one means by useful.

I expect that antivirus drugs are not going to be some miracle cure with respect to novel viruses.

To make effective treatments, scientists need sustained research and experimentation.

Picking a a drug out of a bag of 10 or 20 or 30, and hoping that it would do miracles is a bit foolish and unrealistic.
I don't carry a signature because people are easily offended.
AlphaLess
Posts: 3409
Joined: Fri Sep 29, 2017 11:38 pm
Location: Kentucky

Re: Bogleheads community discussion - Coronavirus

Post by AlphaLess »

folkher0 wrote: Thu Apr 23, 2020 7:30 pm That way you can, if you want, build the model yourself.
A lot of people put their code on github or something.

Might be worth to contact him.
I don't carry a signature because people are easily offended.
rkhusky
Posts: 17764
Joined: Thu Aug 18, 2011 8:09 pm

Re: Bogleheads community discussion - Coronavirus

Post by rkhusky »

AlphaLess wrote: Thu Apr 23, 2020 7:17 pm Quoting:

"The parameters/inputs of this simulator are then learned using machine learning techniques that attempts to minimize the error between the projected outputs and the actual results. We utilize daily deaths data reported by each state/country to forecast future deaths. After some additional validation techniques (to minimize a phenomenon called overfitting), we use the learned parameters to simulate the future and make projections."
That doesn't really say anything. All machine learning algorithms do that, starting with the simplest backprop neural net. So, you really don't know what they are using. I would guess some sort of deep learning, because that is a hot area, is powerful, and there is a lot of code and instruction online that makes it easy to get up and running. But that is a shot in dark, because the site has little specific detail on methods.

edit: It appears that they are using US data to project US deaths into the future, which is essentially what I've been doing. Their projections seem to increase as more deaths are announced, not much different than the simplistic curve fits I've been doing.

It would be more credible if they were using S. Korea or Italy data to predict US deaths. And were trying to predict the peak, rather than the total deaths. It doesn't inspire confidence if the "professionals" can't do much better than I can with a spreadsheet.
folkher0
Posts: 912
Joined: Fri Dec 09, 2016 1:48 pm

Re: Bogleheads community discussion - Coronavirus

Post by folkher0 »

AlphaLess wrote: Thu Apr 23, 2020 7:45 pm
folkher0 wrote: Thu Apr 23, 2020 7:30 pm That way you can, if you want, build the model yourself.
A lot of people put their code on github or something.

Might be worth to contact him.
I guess. I wouldn’t know what to do with it. I have no clue how to do half the stuff he says he did. That’s why it usually takes a team. Not just one guy. Maybe this guy has a team. But if he does he doesn’t give them any cred.

Frankly, I just want him to write a paper in the usual way. Or at least put the model up on the site.

Anyway. It’s a model. I will say before you compare it to another model remember that he’s updating it with new data every day. While that sounds good, if you’re trying to use it to model spread of a virus into the future and compare it to another model doing the same , you should probably compare it to the other model based on the day the other model was run.

It’s not really fair to compare his stuff to their stuff if he’s got 3 more weeks (the most Important weeks) of data.
Last edited by folkher0 on Thu Apr 23, 2020 9:33 pm, edited 1 time in total.
User avatar
knpstr
Posts: 2894
Joined: Thu Nov 20, 2014 7:57 pm
Location: Michigan

Re: Bogleheads community discussion - Coronavirus

Post by knpstr »

What are the odds that the best way forward is accepting immunity through exposure as we go through the upcoming 2nd, 3rd, 4th, nth waves?

Some 70% of the U.S. is under 54 years old.
Very little is needed to make a happy life; it is all within yourself, in your way of thinking. -Marcus Aurelius
Teague
Posts: 2524
Joined: Wed Nov 04, 2015 5:15 pm

Re: Bogleheads community discussion - Coronavirus

Post by Teague »

knpstr wrote: Thu Apr 23, 2020 8:41 pm What are the odds that the best way forward is accepting immunity through exposure as we go through the upcoming 2nd, 3rd, 4th, nth waves?

Some 70% of the U.S. is under 54 years old.
Could you please at least arrange for some ice floes for the rest of us, so we can go the tried and true way?
Semper Augustus
folkher0
Posts: 912
Joined: Fri Dec 09, 2016 1:48 pm

Re: Bogleheads community discussion - Coronavirus

Post by folkher0 »

knpstr wrote: Thu Apr 23, 2020 8:41 pm What are the odds that the best way forward is accepting immunity through exposure as we go through the upcoming 2nd, 3rd, 4th, nth waves?

Some 70% of the U.S. is under 54 years old.
Totally fine.

As long as you're willing to accept the fact if, its anything like what happened to NYC ( which maybe now has an estimated 21% prevalence, and which has some of the highest health care capacity in the country) then that your hospitals and ambulances will be completely overwhelmed and that lots of people are gonna die that could be saved and that young people are gonna end up dead and too including people you may know and love AND THAT THOSE THAT SURVIVE MAY BE CHRONICALLY ILL OXYGEN DEPENDANT AND ON DIALYSIS and all your health care employees are gonna quit or get furloughed and the hospitals are gonna go bankrupt and close and the economy is going to crater anyway under those circumstances and THAT THERE IS NO PROMISE OF DURABLE IMMUNITY...

...why do I bother...
veggivet
Posts: 1480
Joined: Sun Jan 25, 2015 12:07 pm
Location: New England

Re: Bogleheads community discussion - Coronavirus

Post by veggivet »

That plan sounds like a great way to guarantee that more than 70% of the population will be under 54 when all is said and done... :oops:
User avatar
knpstr
Posts: 2894
Joined: Thu Nov 20, 2014 7:57 pm
Location: Michigan

Re: Bogleheads community discussion - Coronavirus

Post by knpstr »

folkher0 wrote: Thu Apr 23, 2020 9:05 pm
knpstr wrote: Thu Apr 23, 2020 8:41 pm What are the odds that the best way forward is accepting immunity through exposure as we go through the upcoming 2nd, 3rd, 4th, nth waves?

Some 70% of the U.S. is under 54 years old.
Totally fine.

As long as you're willing to accept the fact if, its anything like what happened to NYC ( which maybe now has an estimated 21% prevalence, and which has some of the highest health care capacity in the country) then that your hospitals and ambulances will be completely overwhelmed and that lots of people are gonna die that could be saved and that young people are gonna end up dead and too including people you may know and love AND THAT THOSE THAT SURVIVE MAY BE CHRONICALLY ILL OXYGEN DEPENDANT AND ON DIALYSIS and all your health care employees are gonna quit or get furloughed and the hospitals are gonna go bankrupt and close and the economy is going to crater anyway under those circumstances and THAT THERE IS NO PROMISE OF DURABLE IMMUNITY...

...why do I bother...
From my understanding hospitals are going bankrupt now, with the current policies in place. From your previous info it seems like the hospitals are actually helping very little. You've stated no treatments really work well, at all. I believe most people on ventilators are hooked up and you watch them die.

You say no promise of immunity, so if so there is no point in waiting for a vaccine.

The option of humanity staying in their house 90% of their lives isn't viable. So we have to make a move to get on with it.

To be clear, elderly can stay extra cautious as well as anyone going around elderly. Mean/median age of death in MI is 73 and 75.

FWIW it is estimated 265,000,000 will face serious hunger as food shortages are likely. Shutting down the economy kills as well. It isn't a "risk-free" option and people are underestimating the risks of the economic shutdown.
Very little is needed to make a happy life; it is all within yourself, in your way of thinking. -Marcus Aurelius
User avatar
knpstr
Posts: 2894
Joined: Thu Nov 20, 2014 7:57 pm
Location: Michigan

Re: Bogleheads community discussion - Coronavirus

Post by knpstr »

veggivet wrote: Thu Apr 23, 2020 9:11 pm That plan sounds like a great way to guarantee that more than 70% of the population will be under 54 when all is said and done... :oops:
What's your ideal plan?
Wait for the 20-30% effective vaccine?

Unfortunately old people have a tendency to die. Average/median age of death in MI is 73/75 from covid
Very little is needed to make a happy life; it is all within yourself, in your way of thinking. -Marcus Aurelius
veggivet
Posts: 1480
Joined: Sun Jan 25, 2015 12:07 pm
Location: New England

Re: Bogleheads community discussion - Coronavirus

Post by veggivet »

My plan would be for those under 54 to be experimentally infected. The survivors will provide herd immunity for the rest of us. :wink:

Edited to add a dramatic breakthrough in treatment unveiled at today's briefing,,,wait for it...




SUNSHINE! Yes, if you're infected, just go outside in the sun and you'll be cured. No wonder all those people were dying inside hospitals...they were INSIDE!
User avatar
typical.investor
Posts: 5263
Joined: Mon Jun 11, 2018 3:17 am

Re: Bogleheads community discussion - Coronavirus

Post by typical.investor »

knpstr wrote: Thu Apr 23, 2020 9:24 pm
folkher0 wrote: Thu Apr 23, 2020 9:05 pm
knpstr wrote: Thu Apr 23, 2020 8:41 pm What are the odds that the best way forward is accepting immunity through exposure as we go through the upcoming 2nd, 3rd, 4th, nth waves?

Some 70% of the U.S. is under 54 years old.
Totally fine.

As long as you're willing to accept the fact if, its anything like what happened to NYC ( which maybe now has an estimated 21% prevalence, and which has some of the highest health care capacity in the country) then that your hospitals and ambulances will be completely overwhelmed and that lots of people are gonna die that could be saved and that young people are gonna end up dead and too including people you may know and love AND THAT THOSE THAT SURVIVE MAY BE CHRONICALLY ILL OXYGEN DEPENDANT AND ON DIALYSIS and all your health care employees are gonna quit or get furloughed and the hospitals are gonna go bankrupt and close and the economy is going to crater anyway under those circumstances and THAT THERE IS NO PROMISE OF DURABLE IMMUNITY...

...why do I bother...
From my understanding hospitals are going bankrupt now, with the current policies in place. From your previous info it seems like the hospitals are actually helping very little. You've stated no treatments really work well, at all. I believe most people on ventilators are hooked up and you watch them die.

You say no promise of immunity, so if so there is no point in waiting for a vaccine.

The option of humanity staying in their house 90% of their lives isn't viable. So we have to make a move to get on with it.

To be clear, elderly can stay extra cautious as well as anyone going around elderly. Mean/median age of death in MI is 73 and 75.

FWIW it is estimated 265,000,000 will face serious hunger as food shortages are likely. Shutting down the economy kills as well. It isn't a "risk-free" option and people are underestimating the risks of the economic shutdown.
Consider too that obesity was associated with a 4.3-fold increase in hospitalization for those with a BMI between 30 and 40, and a 6.2-fold increase for those with a BMI over 40.

viewtopic.php?f=9&t=310742&p=5203893#p5203856

Last I saw, that was roughly 20%-30% of people 19-55, and 14-20% for children. The 19-55 data was a little old though so might be higher now.

A study in Lille France of patients admitted to intensive care for COVID-19 found almost half were obese or severely obese. Of those who were intubated, nearly 90 percent had a BMI over 35. (published in April 9 Obesity).

I really want the economy opened too. Desperately. But I am not sure throwing that large a percent of vulnerable people under the bus is really humane.

I hope we find a way out soon.
folkher0
Posts: 912
Joined: Fri Dec 09, 2016 1:48 pm

Re: Bogleheads community discussion - Coronavirus

Post by folkher0 »

rkhusky wrote: Thu Apr 23, 2020 8:15 pm
That doesn't really say anything. All machine learning algorithms do that, starting with the simplest backprop neural net. So, you really don't know what they are using. I would guess some sort of deep learning, because that is a hot area, is powerful, and there is a lot of code and instruction online that makes it easy to get up and running. But that is a shot in dark, because the site has little specific detail on methods.

edit: It appears that they are using US data to project US deaths into the future, which is essentially what I've been doing. Their projections seem to increase as more deaths are announced, not much different than the simplistic curve fits I've been doing.

It would be more credible if they were using S. Korea or Italy data to predict US deaths. And were trying to predict the peak, rather than the total deaths. It doesn't inspire confidence if the "professionals" can't do much better than I can with a spreadsheet.
That's external validation.

Also, he doesn't really show us what he does, so I don't know if he's really doing anything more sophisticated than youre modeling. It would be very easy to do what I initially thought cuomo was doing today and look at number of deaths, peak a reasonable cause specific mortality and infer a total number of infected. It would look exactly like his model without all the "machine learning, real time mumbo jumbo"
folkher0
Posts: 912
Joined: Fri Dec 09, 2016 1:48 pm

Re: Bogleheads community discussion - Coronavirus

Post by folkher0 »

knpstr wrote: Thu Apr 23, 2020 9:24 pm From my understanding hospitals are going bankrupt now, with the current policies in place. From your previous info it seems like the hospitals are actually helping very little. You've stated no treatments really work well, at all. I believe most people on ventilators are hooked up and you watch them die.

You say no promise of immunity, so if so there is no point in waiting for a vaccine.

The option of humanity staying in their house 90% of their lives isn't viable. So we have to make a move to get on with it.

To be clear, elderly can stay extra cautious as well as anyone going around elderly. Mean/median age of death in MI is 73 and 75.

FWIW it is estimated 265,000,000 will face serious hunger as food shortages are likely. Shutting down the economy kills as well. It isn't a "risk-free" option and people are underestimating the risks of the economic shutdown.

I will do this once again,

Hospitals won't go bankrupt if we can have enough space to still do normal doctor stuff and not get squashed by COVID.

I don't want the current policy in place. This is unsustainable. That is obvious. Shelter in place is a temporary move that buys you time to make a plan. I want that plan, like immediately. Its my most urgent concern. Don't ask me what plan I want. If I tell you we get nuked by the mods (who are doing an excellent job).

FWIW we estimate survival to discharge for ventilated patients to be 55-70%. Turns out we are better then we thought. That would be much lower if we burst through our capacity. And I've explained before, in response to a previous post someone made stating that doctors were killing patients with ventilators, how and why we use them.

It doesn't have to be like this. Please stop making a false dichotomy between the current untenable situation and mass casualties and the collapse of health care. A reasonable, actionable plan for us to get through this without either outcome is not crazy, and I am personally pretty open minded about what that plan could include.
AlphaLess
Posts: 3409
Joined: Fri Sep 29, 2017 11:38 pm
Location: Kentucky

Re: Bogleheads community discussion - Coronavirus

Post by AlphaLess »

rkhusky wrote: Thu Apr 23, 2020 8:15 pm
AlphaLess wrote: Thu Apr 23, 2020 7:17 pm Quoting:

"The parameters/inputs of this simulator are then learned using machine learning techniques that attempts to minimize the error between the projected outputs and the actual results. We utilize daily deaths data reported by each state/country to forecast future deaths. After some additional validation techniques (to minimize a phenomenon called overfitting), we use the learned parameters to simulate the future and make projections."
That doesn't really say anything. All machine learning algorithms do that, starting with the simplest backprop neural net. So, you really don't know what they are using. I would guess some sort of deep learning, because that is a hot area, is powerful, and there is a lot of code and instruction online that makes it easy to get up and running. But that is a shot in dark, because the site has little specific detail on methods.

edit: It appears that they are using US data to project US deaths into the future, which is essentially what I've been doing. Their projections seem to increase as more deaths are announced, not much different than the simplistic curve fits I've been doing.

It would be more credible if they were using S. Korea or Italy data to predict US deaths. And were trying to predict the peak, rather than the total deaths. It doesn't inspire confidence if the "professionals" can't do much better than I can with a spreadsheet.
Deep learning? Probably overkill. But you said: it is not that hard to setup a colab project, create a simple keras model, etc.

At a high level, they are not doing very different from what you are doing: tracking the hidden value of R0.

But they are simulating an infectious disease model (SEIS), and inferring the parameters of that model: beta, gamma, nu, or whatever other parameters there are.

Using S. Korea or Italy to predict US deaths does not seem like a good exercise. What is a good exercise is to estimate / infer hidden parameters of the SEIS model, which is what they are doing. Those parameters are SPECIFIC to a locale, based on many conditions: e.g, density of a city or state, stage in epidemic, and social distancing measures.
I don't carry a signature because people are easily offended.
AlphaLess
Posts: 3409
Joined: Fri Sep 29, 2017 11:38 pm
Location: Kentucky

Re: Bogleheads community discussion - Coronavirus

Post by AlphaLess »

veggivet wrote: Thu Apr 23, 2020 9:31 pm My plan would be for those under 54 to be experimentally infected. The survivors will provide herd immunity for the rest of us. :wink:

Edited to add a dramatic breakthrough in treatment unveiled at today's briefing,,,wait for it...




SUNSHINE! Yes, if you're infected, just go outside in the sun and you'll be cured. No wonder all those people were dying inside hospitals...they were INSIDE!
I have received a UCV Sanitizer box that has a 15 liter internal capacity. Should have ordered one that has 250 liter internal capacity so that I could fit inside and sanitize myself.

/s
I don't carry a signature because people are easily offended.
AlphaLess
Posts: 3409
Joined: Fri Sep 29, 2017 11:38 pm
Location: Kentucky

Re: Bogleheads community discussion - Coronavirus

Post by AlphaLess »

folkher0 wrote: Thu Apr 23, 2020 8:39 pm
AlphaLess wrote: Thu Apr 23, 2020 7:45 pm
folkher0 wrote: Thu Apr 23, 2020 7:30 pm That way you can, if you want, build the model yourself.
A lot of people put their code on github or something.

Might be worth to contact him.
I guess. I wouldn’t know what to do with it. I have no clue how to do half the stuff he says he did. That’s why it usually takes a team. Not just one guy. Maybe this guy has a team. But if he does he doesn’t give them any cred.

Frankly, I just want him to write a paper in the usual way. Or at least put the model up on the site.

Anyway. It’s a model. I will say before you compare it to another model remember that he’s updating it with new data every day. While that sounds good, if you’re trying to use it to model spread of a virus into the future and compare it to another model doing the same , you should probably compare it to the other model based on the day the other model was run.

It’s not really fair to compare his stuff to their stuff if he’s got 3 more weeks (the most Important weeks) of data.
No offense: but instead of typing here, you can just write him an e-mail.

You are very interested in litigating a problem on your own. Which is fine. But it is not actionable.
I don't carry a signature because people are easily offended.
folkher0
Posts: 912
Joined: Fri Dec 09, 2016 1:48 pm

Re: Bogleheads community discussion - Coronavirus

Post by folkher0 »

AlphaLess wrote: Thu Apr 23, 2020 10:42 pm
folkher0 wrote: Thu Apr 23, 2020 8:39 pm
AlphaLess wrote: Thu Apr 23, 2020 7:45 pm
folkher0 wrote: Thu Apr 23, 2020 7:30 pm That way you can, if you want, build the model yourself.
A lot of people put their code on github or something.

Might be worth to contact him.
I guess. I wouldn’t know what to do with it. I have no clue how to do half the stuff he says he did. That’s why it usually takes a team. Not just one guy. Maybe this guy has a team. But if he does he doesn’t give them any cred.

Frankly, I just want him to write a paper in the usual way. Or at least put the model up on the site.

Anyway. It’s a model. I will say before you compare it to another model remember that he’s updating it with new data every day. While that sounds good, if you’re trying to use it to model spread of a virus into the future and compare it to another model doing the same , you should probably compare it to the other model based on the day the other model was run.

It’s not really fair to compare his stuff to their stuff if he’s got 3 more weeks (the most Important weeks) of data.
No offense: but instead of typing here, you can just write him an e-mail.

You are very interested in litigating a problem on your own. Which is fine. But it is not actionable.
Huh?

You linked to a model. You like the model. You posted it here for discussion. I looked at it. Told you what I thought. Told you why its out of the ordinary for biomedical research. We've had a discussion. That's about it. That's what this place is for, right? Its an anonymous internet forum for personal finance.

I don't really want to pursue it further.

I'm an academic physician. We do things in pretty standard ways. When someone doesn't do it in the standard ways, we ask questions. What else do you want me to say?
AlphaLess
Posts: 3409
Joined: Fri Sep 29, 2017 11:38 pm
Location: Kentucky

Re: Bogleheads community discussion - Coronavirus

Post by AlphaLess »

folkher0 wrote: Thu Apr 23, 2020 10:51 pm
AlphaLess wrote: Thu Apr 23, 2020 10:42 pm
folkher0 wrote: Thu Apr 23, 2020 8:39 pm
AlphaLess wrote: Thu Apr 23, 2020 7:45 pm
folkher0 wrote: Thu Apr 23, 2020 7:30 pm That way you can, if you want, build the model yourself.
A lot of people put their code on github or something.

Might be worth to contact him.
I guess. I wouldn’t know what to do with it. I have no clue how to do half the stuff he says he did. That’s why it usually takes a team. Not just one guy. Maybe this guy has a team. But if he does he doesn’t give them any cred.

Frankly, I just want him to write a paper in the usual way. Or at least put the model up on the site.

Anyway. It’s a model. I will say before you compare it to another model remember that he’s updating it with new data every day. While that sounds good, if you’re trying to use it to model spread of a virus into the future and compare it to another model doing the same , you should probably compare it to the other model based on the day the other model was run.

It’s not really fair to compare his stuff to their stuff if he’s got 3 more weeks (the most Important weeks) of data.
No offense: but instead of typing here, you can just write him an e-mail.

You are very interested in litigating a problem on your own. Which is fine. But it is not actionable.
Huh?

You linked to a model. You like the model. You posted it here for discussion. I looked at it. Told you what I thought. Told you why its out of the ordinary for biomedical research. We've had a discussion. That's about it. That's what this place is for, right? Its an anonymous internet forum for personal finance.

I don't really want to pursue it further.

I'm an academic physician. We do things in pretty standard ways. When someone doesn't do it in the standard ways, we ask questions. What else do you want me to say?
Ok, thanks. Understood. Related to your story with the fantasy football.

A few years back (maybe 2011), my wife was asked to join some type of a football office pool. They send a weekly spreadsheet of the NFL games, and you were supposed to either predict the outright outcome (Win / Loss), or something like that. So, after 3-4 weeks, she complains that she is losing. Mind ya: there are like 50 or 60 people in this pool, and losing is relative. She was close to the bottom.

So I started filling out for her. It took me a week or two. I research football models online (already knew a couple), paid two or three subscriptions (like $10 per), and started doing that every week. The spreadsheet comes out. I use my subscription web-sites, weighted average them, quickly check a few things, and then submit.

In the remaining part of the season my wife's rank went from the bottom tier (like bottom 25%) to 2nd. I could not beat the top person because we were like 3-4 weeks behind. Also, the top 3 or 4 people were REALLY good. I mean, they could make money betting in casinos.

So, there was some prize (like $50) for 2nd place. Then they send this invitation to enter the college-bowl pool. Some phone call comes on Saturday morning or something like that: "Oh, you should enter, blah, blah". And mind ya, there are games that afternoon already. I was really tired of he whole B/S (too much time for too little value), but my wife is like, "Oh, yea, let's enter, we are so good". So we used the winning $50 to enter this pool.

So I have like 2-3 hours to fill out he first game outcomes. Now, what I did in like a week before (find NFL game predictive subscription sites), I had to do in like 2-3 hours. So, off we went.

The tournament was about 4 games short of finishing, and our position was so high, that even if I completely blew the remaining 4 games, we would have won. This is out of 50 people. I can't remember, but I think on the last game (BCS championship), I made an outrageous bet, with a comment like, "You suckers, should not have invited us to enter your pool". I think the prize was like $500 or something.

My point is: I don't spend my time building models in fields that I don't understand well. But I do know enough on how to do an 80-20 to pick the better models. I don't know if this model I linked to is good, great, or awesome. I spent about 5 minutes with the IHME model and realized it was garbage. I spent about 30 minutes with this model, and realized it is NOT garbage.

In my book "Not garbage" >> "garbage".

I am willing to entertain better models any time :D
:sharebeer
I don't carry a signature because people are easily offended.
User avatar
typical.investor
Posts: 5263
Joined: Mon Jun 11, 2018 3:17 am

Re: Bogleheads community discussion - Coronavirus

Post by typical.investor »

As for the suggested injection of disinfectants to counteract COVID-19:
US Centers for Disease Control and Prevention warned Americans to be careful with cleaning products as sales of household disinfectants soar amid the pandemic.

"Calls to poison centres increased sharply at the beginning of March 2020 for exposures to both cleaners and disinfectants," found the agency's weekly morbidity and mortality report.

The US Food and Drug Administration has warned against ingesting disinfectants, citing the sale of bogus miracle cures that contain bleach and purport to treat everything from autism to Aids and hepatitis.

The agency's website says: "The FDA has received reports of consumers who have suffered from severe vomiting, severe diarrhoea, life-threatening low blood pressure caused by dehydration, and acute liver failure after drinking these products."
Locked