3 ms·

> the papers emphasize they are not predictions, but illustrating complex relationships If it has no predictive power, why should it be taken seriously? I'm co

by ahepp 3y ago

> the papers emphasize they are not predictions, but illustrating complex relationships If it has no predictive power, why should it be taken seriously? I'm concerned about the environment and sustainability too, but that's the exact reason we can't afford to throw empiricism out the window the second things start looking a bit hairy.

john-radio 3y ago
A paper not containing predictions is not the same thing as its having no predictive power!
- pclmulqdq 3y ago
  This paper makes a lot of predictions despite the lack of predictive power. Hence, it should not be taken seriously. Since the 1970's the group producing it has made a lot of very wrong predictions out of the hubristic belief that they can reduce everything about life to a simple system of differential equations.
- ahepp 3y ago
  What? Maybe my brain isn't big enough to understand this, but I don't see how you could disentangle the two without getting into pedantry.
  - true_religion 3y ago
    Paper may say you can use this methodology with data beyond the scope of the paper to make a prediction. However it doesn’t actually do the work to predict anything. Kind of like the difference between theoretical math and applied math.
  - john-radio 3y ago
    Take a look, for example, at the timeline in https://xkcd.com/1732/ https://xkcd.com/1732/. It contains (almost) no prediction, but once you have internalized the information in it your own predictions on the climate topic are likely to become a lot more powerful because better informed.
westcort 3y ago
Unfortunately for all of us, the model does have very good predictive power despite its simplicity. Herrington et al in 2020 checked the 1972 model scenarios (there were 4 such scenarios as sensitivity analyses) against around 40 years of empirical data and the results were close to observed ground truth [0]. 0. https://onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fjiec.13084&file=jiec13084-sup-0001-SuppMatS1.pdf https://onlinelibrary.wiley.com/action/downloadSupplement?do...
- Enginerrrd 3y ago
  I try not to be too harsh on here, but your interpretation of this analysis is unbelievably bad. It's absurdly easy to fit a logistic function to match the real-life logistic curve of some value of interest during the exponential phase of those functions. However... That means almost *NOTHING* with regard to the actual logistic portion which is the part people actually care about. There exist an infinite range of different logistic functions that can be made to match in the exponential portion and the majority of those functions contain almost zero information about the future logistic portion. The error between them is essentially unbounded. Calling that confirmation of predictive power is not misguided, it's just about as wrong as you can be.
  - westcort 3y ago
    This is not my analysis, and the first author, Gaya Herrington, explains the alignment between the model and empirical data here: https://m.youtube.com/watch?v=KJpvc3i5j84 https://m.youtube.com/watch?v=KJpvc3i5j84 I also react with concern. Yes there are infinite curves, but there are limits to what is possible, so in actual fact there are not infinitely many possible futures here, and those futures are likely to fall within the limits of the 4 sensitivity analyses…and indeed have over more than 4 decades of follow up.
    - Enginerrrd 3y ago
      I know it's not your analysis. I said your interpretation of this analysis is flawed. You're misunderstanding that analysis to be confirmation of the predictive power of the model extrapolated into the logistic region. And you still are: >Yes there are infinite curves, but there are limits to what is possible, so in actual fact there are not infinitely many possible futures here, and those futures are likely to fall within the limits of the 4 sensitivity analyses…and indeed have over more than 4 decades of follow up. The key phrase there where your interpretation is flawed is: "are likely to fall within the limits of the 4 sensitivity analyses". That's so, so untrue. You simply cannot extrapolate logistic model parameters into the actual post-exponential-growth, logistic portion without seeing the actual logistic effects start to kick in. Those simply haven't been empirically verified here. And speaking of... the original model they are analyzing introduces limits on that exponential growth in a pretty arbitrary and artificial way, particularly when it comes to the pollution terms.... which, I'll add, this analysis shows is also the parameter they most over-estimated by a large amount even in the exponential portion.

by ahepp 3y ago

john-radio 3y ago

A paper not containing predictions is not the same thing as its having no predictive power!

pclmulqdq 3y ago
This paper makes a lot of predictions despite the lack of predictive power. Hence, it should not be taken seriously. Since the 1970's the group producing it has made a lot of very wrong predictions out of the hubristic belief that they can reduce everything about life to a simple system of differential equations.
ahepp 3y ago
What? Maybe my brain isn't big enough to understand this, but I don't see how you could disentangle the two without getting into pedantry.
- true_religion 3y ago
  Paper may say you can use this methodology with data beyond the scope of the paper to make a prediction. However it doesn’t actually do the work to predict anything. Kind of like the difference between theoretical math and applied math.
- john-radio 3y ago
  Take a look, for example, at the timeline in https://xkcd.com/1732/ https://xkcd.com/1732/. It contains (almost) no prediction, but once you have internalized the information in it your own predictions on the climate topic are likely to become a lot more powerful because better informed.

westcort 3y ago

Unfortunately for all of us, the model does have very good predictive power despite its simplicity. Herrington et al in 2020 checked the 1972 model scenarios (there were 4 such scenarios as sensitivity analyses) against around 40 years of empirical data and the results were close to observed ground truth [0]. 0. https://onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fjiec.13084&file=jiec13084-sup-0001-SuppMatS1.pdf https://onlinelibrary.wiley.com/action/downloadSupplement?do...

Enginerrrd 3y ago
I try not to be too harsh on here, but your interpretation of this analysis is unbelievably bad. It's absurdly easy to fit a logistic function to match the real-life logistic curve of some value of interest during the exponential phase of those functions. However... That means almost *NOTHING* with regard to the actual logistic portion which is the part people actually care about. There exist an infinite range of different logistic functions that can be made to match in the exponential portion and the majority of those functions contain almost zero information about the future logistic portion. The error between them is essentially unbounded. Calling that confirmation of predictive power is not misguided, it's just about as wrong as you can be.
- westcort 3y ago
  This is not my analysis, and the first author, Gaya Herrington, explains the alignment between the model and empirical data here: https://m.youtube.com/watch?v=KJpvc3i5j84 https://m.youtube.com/watch?v=KJpvc3i5j84 I also react with concern. Yes there are infinite curves, but there are limits to what is possible, so in actual fact there are not infinitely many possible futures here, and those futures are likely to fall within the limits of the 4 sensitivity analyses…and indeed have over more than 4 decades of follow up.
  - Enginerrrd 3y ago
    I know it's not your analysis. I said your interpretation of this analysis is flawed. You're misunderstanding that analysis to be confirmation of the predictive power of the model extrapolated into the logistic region. And you still are: >Yes there are infinite curves, but there are limits to what is possible, so in actual fact there are not infinitely many possible futures here, and those futures are likely to fall within the limits of the 4 sensitivity analyses…and indeed have over more than 4 decades of follow up. The key phrase there where your interpretation is flawed is: "are likely to fall within the limits of the 4 sensitivity analyses". That's so, so untrue. You simply cannot extrapolate logistic model parameters into the actual post-exponential-growth, logistic portion without seeing the actual logistic effects start to kick in. Those simply haven't been empirically verified here. And speaking of... the original model they are analyzing introduces limits on that exponential growth in a pretty arbitrary and artificial way, particularly when it comes to the pollution terms.... which, I'll add, this analysis shows is also the parameter they most over-estimated by a large amount even in the exponential portion.