3 ms·

There's some minor variation due to different curve-fitting approaches, but the big variation (e.g. R0 from 3.7–203.3 for measles, per my other comment here) is

There's some minor variation due to different curve-fitting approaches, but the big variation (e.g. R0 from 3.7–203.3 for measles, per my other comment here) is real, simply because the environment or human behavior varied. As others have repeatedly noted, R0 is a function not just of the pathogen, but also of its environment, including the behavior of its human hosts. Earlier, you wrote: > It [R0] was never being established via lab work, as you might expect given its definition. If you expected that "lab work" could establish R0, then you've grossly misunderstood its meaning and definition. It seems like you're looking for a concept of "R0 but for the pathogen alone, independent of environment and human behavior". That just doesn't exist though, any more than you could define the growth rate of a plant independent of weather and soil fertility.

mike_hearn 3y ago
Briefly, as the substance is discussed in my other reply: > the big variation (e.g. R0 from 3.7–203.3 for measles, per my other comment here) is real Morpheus: "What is real? How do you define, real?" CIs that wide are just an obfuscated way of saying "we have no idea what's going on or what will happen". Anyone can make predictions that way. For example by the end my life my bank balance will be $4.7 million (CI $5.00-$50M). Those numbers aren't "real" in any meaningful sense. Anyone can express don't-know in sophisticated looking numerical form, and they wouldn't justify me claiming to be a financial expert on the back of them. > any more than you could define the growth rate of a plant independent of weather and soil fertility. You can define growth rate of a plant this way: define a standardized lab environment with regularized soil composition and artificial light/watering schedules. Then plant seeds and measure the dry weight at the end of a fixed time period. This will give you a number that's comparable across species. There are other definitions you could use because "growth rate" is slightly ambiguous in English (does a tree grow faster than a weed because the tree achieves bigger mass?), but that's the general idea. What you shouldn't do is just grab datasets of wildly varying quality off the internet, fit an equation you just invented on the spot to that data, and announce you've discovered something real about the plants in the data. That would indeed yield a growth rate that doesn't tell you anything meaningful.
- tripletao 3y ago
  > You can define growth rate of a plant this way: define a standardized lab environment [...] If the standardized lab environment is dry and sunny, then you'll conclude that a cactus grows faster than a fern, since the ferns will mostly shrivel up and die. If the standardized lab environment is moist and shaded, then you'll conclude that the fern grows faster, since cactuses will mostly die for lack of sun. So which is right? The concept that you're looking for simply doesn't exist--the growth rate of an organism can't be defined except with reference to its environment, which for a virus that infects humans includes human behavior. (What rate of condom use should the standardized lab environment for HIV correspond to? How will you model the increased popularity of fentanyl?) You are looking for CS-level rigor and simplicity in biology, but biology doesn't work like that. You are correct that many biological results were thus oversold to the public during the pandemic; but you're once again criticizing those public-facing oversimplifications, not any science as a practitioner would understand it.
  - mike_hearn 3y ago
    Neither can be said to be right without reference to a fixed context. If your field standardizes on a lab setup that's drier and sunnier than what ferns like, you'd indeed conclude that ferns grow slower in that context and that's OK because the growth rate is at least well defined. If there's a use case for comparable growth rates in different contexts, OK, define separate names for those rates and measure them separately. Or try to isolate the effect of heat and light such that the growth rate of any plant can be computed from the equivalent of e=mc^2. Likewise you wouldn't try to measure the infectiousness of HIV in people, clearly. If you want to measure the relative "infectiousness" of viruses in humans using precise numbers then you'd need a controlled experimental environment, presumably something in vitro. That would miss a lot of factors that are important if you're trying to predict epidemics at the society-wide level, but OK, so be it. You need a firm footing of the basics before you can progress to more complex scenarios. I don't really agree that it's unreasonable to expect CS-level rigor in biology. Microbiologists seem to manage it? It's expected that if two labs sequence the same organism they can in principle get the same DNA sequence, and if they do Xray crystallography on the same protein they'll derive the same structure. So we're not even comparing biology and CS here, we're comparing microbiology with epidemiology. The latter seems to be far closer to a social science in terms of its methods and rigor. To be clear, it's also fine to do epidemiology using less rigorous methods if it was done in the way it mostly used to be done. When I read papers from the 80s they seemed to be much more appropriate to the actual data quality - largely prose oriented, very limited use of maths, presenting falsifiable hypotheses whilst admitting to the big unknowns. That's fine, science doesn't always have to be precisely quantifiable especially on the margins of what's known. But if scientists do precisely quantify things, then those quantities should be well defined.

mike_hearn 3y ago

Briefly, as the substance is discussed in my other reply: > the big variation (e.g. R0 from 3.7–203.3 for measles, per my other comment here) is real Morpheus: "What is real? How do you define, real?" CIs that wide are just an obfuscated way of saying "we have no idea what's going on or what will happen". Anyone can make predictions that way. For example by the end my life my bank balance will be $4.7 million (CI $5.00-$50M). Those numbers aren't "real" in any meaningful sense. Anyone can express don't-know in sophisticated looking numerical form, and they wouldn't justify me claiming to be a financial expert on the back of them. > any more than you could define the growth rate of a plant independent of weather and soil fertility. You can define growth rate of a plant this way: define a standardized lab environment with regularized soil composition and artificial light/watering schedules. Then plant seeds and measure the dry weight at the end of a fixed time period. This will give you a number that's comparable across species. There are other definitions you could use because "growth rate" is slightly ambiguous in English (does a tree grow faster than a weed because the tree achieves bigger mass?), but that's the general idea. What you shouldn't do is just grab datasets of wildly varying quality off the internet, fit an equation you just invented on the spot to that data, and announce you've discovered something real about the plants in the data. That would indeed yield a growth rate that doesn't tell you anything meaningful.

tripletao 3y ago
> You can define growth rate of a plant this way: define a standardized lab environment [...] If the standardized lab environment is dry and sunny, then you'll conclude that a cactus grows faster than a fern, since the ferns will mostly shrivel up and die. If the standardized lab environment is moist and shaded, then you'll conclude that the fern grows faster, since cactuses will mostly die for lack of sun. So which is right? The concept that you're looking for simply doesn't exist--the growth rate of an organism can't be defined except with reference to its environment, which for a virus that infects humans includes human behavior. (What rate of condom use should the standardized lab environment for HIV correspond to? How will you model the increased popularity of fentanyl?) You are looking for CS-level rigor and simplicity in biology, but biology doesn't work like that. You are correct that many biological results were thus oversold to the public during the pandemic; but you're once again criticizing those public-facing oversimplifications, not any science as a practitioner would understand it.
- mike_hearn 3y ago
  Neither can be said to be right without reference to a fixed context. If your field standardizes on a lab setup that's drier and sunnier than what ferns like, you'd indeed conclude that ferns grow slower in that context and that's OK because the growth rate is at least well defined. If there's a use case for comparable growth rates in different contexts, OK, define separate names for those rates and measure them separately. Or try to isolate the effect of heat and light such that the growth rate of any plant can be computed from the equivalent of e=mc^2. Likewise you wouldn't try to measure the infectiousness of HIV in people, clearly. If you want to measure the relative "infectiousness" of viruses in humans using precise numbers then you'd need a controlled experimental environment, presumably something in vitro. That would miss a lot of factors that are important if you're trying to predict epidemics at the society-wide level, but OK, so be it. You need a firm footing of the basics before you can progress to more complex scenarios. I don't really agree that it's unreasonable to expect CS-level rigor in biology. Microbiologists seem to manage it? It's expected that if two labs sequence the same organism they can in principle get the same DNA sequence, and if they do Xray crystallography on the same protein they'll derive the same structure. So we're not even comparing biology and CS here, we're comparing microbiology with epidemiology. The latter seems to be far closer to a social science in terms of its methods and rigor. To be clear, it's also fine to do epidemiology using less rigorous methods if it was done in the way it mostly used to be done. When I read papers from the 80s they seemed to be much more appropriate to the actual data quality - largely prose oriented, very limited use of maths, presenting falsifiable hypotheses whilst admitting to the big unknowns. That's fine, science doesn't always have to be precisely quantifiable especially on the margins of what's known. But if scientists do precisely quantify things, then those quantities should be well defined.