COVID-19

Health, Fitness, Insurance, ...
steveo73
Posts: 1432
Joined: Sat Jul 06, 2013 6:52 pm

Re: COVID-19

Post by steveo73 »

bigato wrote:
Thu May 21, 2020 7:32 pm
steveo: if you really think any prediction made here or elsewhere was made with certainty, I'm very sorry that you misunderstood that so badly.
Let's be very clear in that I haven't changed my position throughout this discussion. I have never stated that predictions have any certainty at all attached to them. They are based on scientific models and these models ala models of how to trade financial instruments are fraught with problems. It is in their very nature.

I'm articulating the same issue that you seem to be agreeing with me here about so I guess that is progress. Statistical models may or may not represent reality well enough to model the behavior they are attempting to model but these models are fraught with danger. Science is not so clear cut a lot of the time that we can use simplistic models and state this is reality. It isn't. The certainty shouldn't exist.

The rest of your points I've been through. It's very very clear that we are not clear on how this virus will progress throughout the world and I'm good with that. There will be multiple reasons for that. It's really really hard to model complex scenarios.

I think the outcome should be really clear though - going forward let's not use these models and state it's the science or it's the math as if it's going to model the future accurately. Let's start using a risk management framework to manage and discuss these types of issues and let's be honest and open about the science.

jacob
Site Admin
Posts: 12420
Joined: Fri Jun 28, 2013 8:38 pm
Location: USA, Zone 5b, Koppen Dfa, Elev. 620ft, Walkscore 73
Contact:

Re: COVID-19

Post by jacob »

WIRED magazine did an interesting series where experts explains and talks about something (CRISPR, sleep, gravity, harmonies, lasers, ... but unfortunately not epidemilogy) at 5 different levels of complexity, respectively, a child, a teen, a college student, a grad student, and another expert.

https://www.youtube.com/playlist?list=P ... npAIlKgm3N

I'd recommend listening to one in a field you're not familiar with as well as one in the subject you know the most about and watch the dynamics of the conversation as well as your own reactions.

Some hopeful takeaways from watching a couple of episodes is first that in most subjects you're not familiar with, you probably know as much as the average teen but not much more. You can still follow some or most of the college student discussion, but some info/words/concepts are unfamiliar. The grad and expert level is likely beyond you and it all begins to sound like techno babble---you could not be an equal contributory participant at those levels. Second, the video of the subject you know the most about seems strangely simple in comparison to the others and you find yourself wondering why they didn't pick harder levels of details. This is of course due to the curse of knowledge and not knowing how much of your knowledge you take for granted. This in turn is a challenge when trying to explain yourself to other nonexperts.

When issuing public statements (on TV), experts find themselves talking to the child/teen level because that's where most of the adult audience is---most people don't study further than HS outside their specialization. If Fauci showed up on PBS or in the Senate and gave a full professional account, most of the Senate and TV audience would hear it as techno babble. Of course these are not the kind of discussions he's having with his peers. See the WIRED videos to see what I mean.

I'd say (and this is mostly to steveo73 because everybody else seems to get it) that if anyone thinks that experts are using simplistic models without realizing it just because that's how they're talking about simple models to explain their points, they need to change the channel and stop getting their information from blog posts, twitter, or t-shirt slogans. Constantly complaining about the lack of details while never going into any details yourself beyond dismissing all predictions as alarmists, poor science, fraught with danger, or <gasp> "exponential" is just tilting at windmills. Wanting to generalize this attitude willy-nilly does indeed sound like a crusade against any kind of prediction whatsoever.

To have expert-to-expert conversations or expert-level information about all the uncertainties and complications at a level that is way above the paygrade of the average armchair tweeter, go read the actual papers being put out. Before doing that, go read a textbook or several so you have some framework for understanding what you're reading.

daylen
Posts: 1584
Joined: Wed Dec 16, 2015 4:17 am

Re: COVID-19

Post by daylen »

To add to what Jacob said, a bit of empathy can go a long way. Immediate dismissal of an argument based entirely on your own idiosyncratic parsing is often not a good indicator of disagreement (communication has many points of failure). People use words in different ways. It is often much easier to work with people if you allow for the possibility that their definitions [..and ideally priors] are different from yours. A good exercise to see if you understand someone else is to switch out a few keywords they use with placeholders. This will reveal to underlying structure and can imply the structure of a reasonable response(*), then try to imagine the range of intended meanings for each keyword. This requires practice and a genuine interest in what others have to say.

Take "science" as an example. A more exclusive definition may be synonymous with "objective truth" or just "truth". A more inclusive definition may be an institution [of people, tools, and agendas] that generates hypotheses to be validated with data fitted to models. Even more inclusive would recognize that industry often drives the process of agenda selection and that multiple groups of ideologically polarized researchers can falsify each other (incidentally correcting for overconfidence and sampling biases). Multiple studies can be analyzed in a meta-study to validate the implementation of statistical methodology.

Reality is complex, but the process of science is too [by necessity]. From a different angle, it is straightforward to make the case that prediction is feasible in a wide array of situations (there is plenty of evidence if you are willing to look and attempt understanding). Even the most unpredictable range of objects (organs to climates) are constrained enough for models to be used by millions of people every day. Precision =/= Accuracy; science is typically accurate with a lack of precision; math is an efficient language for precision but can handle accuracy just fine(**); lack of scientific precision does not imply lack of accuracy or lack of utility.

If several others do not accept your view and you do not understand why, then learning is possible. Learning may require you to revisit your most durable assumptions and potentially dismantle how you previously made sense of reality. The resurrection is often sweet, though.

(*) Until this process becomes second nature, of course. Parsing is a skill too. For a more advanced rabbit hole look into phrase structure and how it can alter interpretation or just tire the reader if done wrong.

(**) Precise statement: I am 5.61785 feet tall. Accurate statement: I am 5'7" give or take 0.5 inches. The precise statement indicates measurements hit that bullseye every time. The accurate statement indicates a cluster of measurements roughly within an inch of each other centered on 5'7".
Last edited by daylen on Fri May 22, 2020 1:17 am, edited 1 time in total.

horsewoman
Posts: 425
Joined: Fri Jun 07, 2019 4:11 am

Re: COVID-19

Post by horsewoman »

@jacob - This is timely, as I experienced such a "level shift" yesterday during one of the Corona-updates of Prof. Drosten. He tries to explain cutting edge science in a way that HS level listeners will get it somewhat - or at least leave the podcast better informed and less prone to spread/believe misinformation. But after "dumbing himself down" for over 30 minutes he started to slip towards the end for the last 15 minutes, his language gradually changed and he used a lot of technical terms and concepts. I only understood most of these because they are derived from Latin words, which in kind are words used in English, not so much in conversational German. It apparently takes a lot of concentration to stay consistently on a lower level!

nomadscientist
Posts: 176
Joined: Fri Mar 13, 2020 12:54 am

Re: COVID-19

Post by nomadscientist »

Are epidemiology models science? Against what experiments are they tested? This isnt a personal criticism as it's not like you can unleash test pandemics on a whim. But it does give grounds for lower confidence than in rocket science, rockets having been tested many times, with most of the early tests failing.

daylen
Posts: 1584
Joined: Wed Dec 16, 2015 4:17 am

Re: COVID-19

Post by daylen »

@nomad There are a variety of techniques that can be employed to gain a more clear view of reality. Anything imaginable can be [temporarily] constrained to some degree within the purview of an observer. Assuming a narrow interpretation of how truth can be discovered/possessed/communicated stunts growth. Yet, too wide of an interpretation is stifling, so there is some balancing to do.

See "Types of studies" section of the "Epidemiology" wikipedia article.

bostonimproper
Posts: 213
Joined: Sun Jul 01, 2018 11:45 am

Re: COVID-19

Post by bostonimproper »

Interesting video from The Atlantic with interviews from COVID-19 survivors who suffered ARDS, severe delirium, and ongoing PTSD symptoms. Definitely worth a watch to get a glimpse into the long-term mental health issues documented with SARS play out in real time with COVID-19.

ertyu
Posts: 935
Joined: Sun Nov 13, 2016 2:31 am

Re: COVID-19

Post by ertyu »

Thank you for sharing this, @bi. Humbling. Also a reminder to honor the reality of covid in the face of the efforts by powers that be to use the illness to fuel a culture war.

User avatar
Sclass
Posts: 1933
Joined: Tue Jul 10, 2012 5:15 pm
Location: Orange County, CA

Re: COVID-19

Post by Sclass »

Oops, that didn’t post well. I’ll Try it again.
jacob wrote:
Thu May 21, 2020 7:51 pm
An example of a model that was not based on science would be a Kalman filter (I presume you're familiar---if not, it's a purely mathematical construct for updating a variable with some prescription under uncertainty ...
Always impressed with the breadth of your knowledge. Good illustrative example too. That is the easiest to understand example of a Kalman filter I’ve seen.

Reminded me that none of the models seem to have an adaptive feedback term for the effect of building fear. We mentioned predator prey a few hundred messages ago, but there should also be coefficients for people freaking out, scrambling to sew masks, refraining from hookups, stop seeing grandma etc. based on their current reckoning of the situation...which is changing as often as the information on Good Morning America. The characteristic peaking of the death rates around the world seem to be influenced by lockdowns where the masses have time to digest the severity of the situation while they’re told to sit down while authority slaps their faces repeatedly.

I’m actually surprised the models fit the data so well. Seriously, anyone here tried to fit real data with a mathematical model? Without a boatload of terms to selectively throw out and some twiddle factors I never could fit anything I measured. Except the boundary and initial conditions which were measured. Yeah conditioning the model is a big job too. I think if things worked as well as they do in the physics 101 lab you earn a Nobel prize or equivalent.
Last edited by Sclass on Fri May 22, 2020 7:34 am, edited 1 time in total.

IlliniDave
Posts: 2959
Joined: Wed Apr 02, 2014 7:46 pm

Re: COVID-19

Post by IlliniDave »

jacob wrote:
Thu May 21, 2020 7:51 pm
@iDave - From the perspective of an ex-physical scientist and ex-financial modeller, I think it's fair to say that "all science is based on models but not all models are based on science". For example, suppose I wanted to build a tracker-predictor, for example, a targeting radar for tracking falling objects.

An example of a model that was not based on science would be a Kalman filter ...
jacob,

The first two sentences, while probably not the way I'd say it, I'd basically agree with. Over the last decade or so I started hearing the term "The Science" invoked a lot. When I hear that I tend to assume what is meant is the underlying "truth" that little-s science (to me, an activity, not a thing) seeks to understand. Mathematics is a language we invented to describe how nature/the universe work. Models of the type discussed here are largely applications of the math. It seems that insofar as we can record, share/discuss, and build on science we are completely reliant on math and a little bit of jargon, and in that sense I can see a certain inseparability. Gives me some new topics for navel gazing. :)

Yes, I've got a working understanding of Kalman Filters, not so sharp now as it was 20 years ago when I spent most of my time slogging through the weeds and my graduate studies were fresh. You probably recall we discussed the topic offline a while back. Kalman was an EE, as am I as far as academic training. Luckily my jargon is rusty so I'll avoid it as much as I can lest I embarrass myself by getting some of it wrong. The way I learned to derive the "Kalman Filter Equation" started with a very simple signal processing problem (extracting signal from observations corrupted by static white noise) then generalized to more complicated systems. My concentration was control systems so from there we took it in a certain direction. Despite that, and that I'm not a radar expert (I can spell it both forwards and backwards though) I'm a little familiar with the tracking of ballistic objects problem. At least as pertains to this planet, ballistics are pretty well understood, and at the heart of such a Kalman filter is a "model" of a falling object (which afaik would include gravity and drag, btw, so I guess my experience is with the expert level, ha). If we're talking about tracking an asteroid or something like that to predict it's impact point, worrying about gravity and drag is pointless because you won't have time get enough updates for the filter to "converge" (Kalman filters, at least of this type, work in "real time" and don't "look ahead" very far), and likely the object is moving so fast the cumulative effects of gravity and drag will be minimal. If you're tracking a satellite whose orbit is decaying, those effects might be more germane. What's not known at first encounter is the object's initial state (position and velocity vector, maybe accelleration) and its size/shape (how drag will affect it). The problem is to sort through inherently noisy and somewhat imprecise radar measurements, and the Kalman filter is a practical computational algorithm to come up with an "optimal"* formulation of the model (calibrate the model you might say). So from my limited perspective I wouldn't classify a Kalman Filter itself as a model, although including a model and predicting ahead incrementally are part of the algorithm. I suppose the model can be as sciencey or unsciencey as the situation warrants.

One way we commonly see modeling that isn't derived from science is when random variables are substituted for phenomena we can observe but either don't understand enough to model properly or the underlying process is prohibitively complex. Sticking with radio waves, antennas pick up "atmospheric noise". We probably have a pretty good understanding of a lot of the processes that produce radio waves across the universe, but detailed modeling of the radio band emission sources of the entire universe to test via modeling (i.e., by simulation) how well a new gps receiver design will reject noise and give accurate readings isn't practical. We have lots of measurements of this noise so we can come up with a suitable representation by matching statistical properties in a random variable. Where it gets more dicey is when we don't have the observation base and use random variables to hopefully cover a wide range of excursion of some unknown. That might be at the heart of the think tank's criticism that 7Wb5 mentioned above.

Your last sentence: "Fundamentally, models are thus ways of describing reality and scientific models are models that have been subject to the scientific method." sums up what I was getting at more expertly than what I did. Which segues to one one my favorite sayings (attributed to Einstein, perhaps apocryphally, I dunno): "As far as the laws of mathematics refer to reality, they are not certain, and as far as they are certain, they do not refer to reality." You could probably substitute "model" for "mathematics" in that statement with no loss of generality. And that's probably what's at the heart of the problem getting hashed out here. Models are just models. Models not subject to the scientific method (yet) need to be understood with caution. Some of the models I've worked on and with were extremely expensive to develop, took years and input from SMEs across many disciplines, and even more money/effort was spent to verify/validate them. Only then were they employed for prediction. I think the problem here isn't so much the initial rough-cut model, but lies either with those who promulgated the results to laypersons, or what the laypersons decided to do with them. And it's hard to even use the word "fault" because I like to believe that somewhere below all the noise people were doing their best to save lives without making the cure more deadly than the disease, as it were.

*To be "optimal" requires some assumptions in the derivation that don't always apply IRL, iirc, but the algorithm still works well enough to be useful.
Last edited by IlliniDave on Fri May 22, 2020 7:47 am, edited 2 times in total.

User avatar
Sclass
Posts: 1933
Joined: Tue Jul 10, 2012 5:15 pm
Location: Orange County, CA

Re: COVID-19

Post by Sclass »

@ID

You sound like you’re from that big rocket lab in Huntsville.

IlliniDave
Posts: 2959
Joined: Wed Apr 02, 2014 7:46 pm

Re: COVID-19

Post by IlliniDave »

Sclass, not exactly, but as it happens I've rubbed elbows with a fair number of those folks over the years.

User avatar
jennypenny
Posts: 6521
Joined: Sun Jul 03, 2011 2:20 pm
Location: Stepford USA

Re: COVID-19

Post by jennypenny »

NatGeo: Inflamed brains, toe rashes, strokes: Why COVID-19's weirdest symptoms are only emerging now

I know four people in their 30s who've had surprise heart attacks in the last month. Three of them died. All of them lived with someone who had COVID. I realize that might be statistical noise since I live in the NY/NJ orbit and we're at the point where 'everybody knows somebody ...', but it's still disturbing (and very sad).

I'm not arguing for or against ending the lockdown (moot in most places anyway). It's more a morbid fascination with the disease and how hard it is to pin down, even six months in. Makes me long for an old-fashioned influenza pandemic.

IlliniDave
Posts: 2959
Joined: Wed Apr 02, 2014 7:46 pm

Re: COVID-19

Post by IlliniDave »

jennypenny wrote:
Fri May 22, 2020 8:36 am
NatGeo: Inflamed brains, toe rashes, strokes: Why COVID-19's weirdest symptoms are only emerging now

I know four people in their 30s who've had surprise heart attacks in the last month. Three of them died. All of them lived with someone who had COVID. I realize that might be statistical noise since I live in the NY/NJ orbit and we're at the point where 'everybody knows somebody ...', but it's still disturbing (and very sad).

I'm not arguing for or against ending the lockdown (moot in most places anyway). It's more a morbid fascination with the disease and how hard it is to pin down, even six months in. Makes me long for an old-fashioned influenza pandemic.
Haven't checked out the link yet, but that's something Weinstein touched on in the link I included above. This virus doesn't really fit the mold. Respiratory stress for some, extreme blood clotting/strokes for others, still others with severe cardiac issues, then the miscellaneous things like "covid toe" and loss of the sense of smell. Almost like it's a "smart" virus that assesses and attacks selectively rather than just mindlessly doing its thing. Maybe my surprise is rooted in ignorance of virus M.O.s, but this thing seems like something out of a science fiction story, getting more spooky as time goes on.

daylen
Posts: 1584
Joined: Wed Dec 16, 2015 4:17 am

Re: COVID-19

Post by daylen »

IlliniDave wrote:
Fri May 22, 2020 7:32 am
Mathematics is a language we invented to describe how nature/the universe work. Models of the type discussed here are largely applications of the math. It seems that insofar as we can record, share/discuss, and build on science we are completely reliant on math and a little bit of jargon, and in that sense I can see a certain inseparability.
Hmm, I was under the impression that early math (arithmetic) was invented to track debts/credits (tributes probably preceded currency). Algebra was largely invented to understand geometry without much involvement in application (not sure about the Babyloian's or Egyptian's). Interest in application exploded around the time of the industrial revolution which then paved the way for more serious scientific institutions (grounded with engineering). Calculus is really the first clear example, I think, of math invented for the purpose of describing natural phenomena. Newton the INTJ went from observation -> description, but Leibniz the INTP concurrently went from idealized problems -> syntax.

Roughly, the domain of analysis evolved to describe/handle observations, and the domain of algebra evolved to solve idealized problems via syntactical manipulation. Yet, analysis and algebra have a great deal of mutuality. They can be considered united just as "activities" must be coupled to "things".

Some fields such as physics are more reliant on math but there are less precise and more observational fields which are not (evolutionary biology for instance). Weinstein has commented on how he finds math to often be misleading. I find it subpar at describing visual patterns (relative to analogy). Technically, English can be converted into Math (along with the inverse), but this may or may not be practical in various situations. Humans are adapted primarily to vocal communication which is preferred when exchanging crude explanations and anecdotal evidence. Brainstorming snowballs into more rigorous methodology, so it is somewhat moot to delineate too strictly between "science" and "not science".

Jason
Posts: 2761
Joined: Mon Jan 30, 2017 8:37 am

Re: COVID-19

Post by Jason »

jennypenny wrote:
Fri May 22, 2020 8:36 am
I know four people in their 30s who've had surprise heart attacks in the last month.
Is a heart attack ever not a surprise? I know some people may be more susceptible than others but does anyone say "Oh, gee, looks like I'm having that heart attack I was expecting."

User avatar
jennypenny
Posts: 6521
Joined: Sun Jul 03, 2011 2:20 pm
Location: Stepford USA

Re: COVID-19

Post by jennypenny »

Redd Foxx?

7Wannabe5
Posts: 6114
Joined: Fri Oct 18, 2013 9:03 am
Location: Clinton River Watershed

Re: COVID-19

Post by 7Wannabe5 »

Anybody who ever said “I’ll have the Hillbilly Pancakes, doublestack.”

Jason
Posts: 2761
Joined: Mon Jan 30, 2017 8:37 am

Re: COVID-19

Post by Jason »

How many of the four had their heart attack on a day other than Monday? You really can't express complete surprise if you're having a heart attack on Monday morning.

7Wannabe5
Posts: 6114
Joined: Fri Oct 18, 2013 9:03 am
Location: Clinton River Watershed

Re: COVID-19

Post by 7Wannabe5 »

Here’s a recent open access article published in the Journal of Mathematical Biology. It’s not on the topic of Covid, but it gives some idea of the level of sophistication somebody like Neal Ferguson would be working on vs. you/me.

https://link.springer.com/article/10.10 ... 19-01351-6

Post Reply