4 ms·
> But if f is NOT linear, the slope will be higher in some places than LNT would imply (this again is a simple theorem from calculus). What does that even mean
by roenxi 7y ago
> But if f is NOT linear, the slope will be higher in some places than LNT would imply (this again is a simple theorem from calculus).
What does that even mean? That is junk maths, I don't believe there is such a theory. A line can have any gradient and I can always come up with an arbitrary curve that has a lower or equal gradient at every point that matches the line.
A model that matches LNT exactly to some threshold then the risk drops to 0, for example. Not higher than the LNT anywhere in any negative sense.
- Retric 7y agoSuppose a vending machine gives 20 soda for 20$. If someone walks up and inserts 1$ and can get 0 soda, but if they insert an extra 19$ they get 20 soda. If that’s true then for some amount between 1$ and 20$ inserting an extra 1$ must let you get more than 1 soda in order for it to be handing out 20 soda at 20$. PS: Back to cancer, if it’s 19 cancer at 20x then you can keep a 1:1 relationship at 2+x and 0 cancer at 1x. Then again if you can’t tell if it’s 19 cancer or 20 cancer then it might just be 21 cancer.
- pfdietz 7y ago> What does that even mean? That is junk maths, It's a consequence of the Mean Value Theorem. Suppose a differentiable function f has f(0) = 0 and f(1) = 1. Then, either f is linear (f(x) = x), or there is some point y for which f(y) != y. If f(y) > y, then by the MVT there must be some point z betwenn 0 and y for which f'(z) >= f(y)/y > 1. If f(y) < y, then again by the MVT there is some point z between y and 1 for which f'(z) > (1 - f(y))/(1 - y) > 1.
- roenxi 7y agoThat isn't making a sensible point. f(x) is the risk at dose x. If f(x) = x under the LNT and g(x) = {0: x<0.5, x: x>0.5} then the Mean Value Theorum is satisfied and the non-linear g() is equal to or less risky than the LNT for any dose. g(x) is a linear-with-threshold model by the way. The steeper slope (infinte, in fact) found in the model would be a huge theoretical positive for nuclear.
- pfdietz 7y agoOF course it's making a sensible point. If the dose response function is not linear (and is a differentiable function, as any physically real function must be) then it will have a point where the derivative is greater than for LNT. Your example is discontinuous. This is not physically realistic. Any "real" function, describing the response of a population with random individual exposures and differences, will be smoothed, and so be better behaved. It also misses the point that when you abandon LNT, you don't also get to say "and now we'll assume the function is even more favorable to my political agenda". Sure, it's possible the actual function makes it better for nuclear advocates. But it's also possible it makes it worse. Are the policy makers going to be on your side just because?
- roenxi 7y ago> Your example is discontinuous. This is not physically realistic. Any "real" function, describing the response of a population with random individual exposures and differences, will be smoothed, and so be better behaved. Eh. If you like. If you use a continuous analogue the argument doesn't change at all. > It also misses the point that when you abandon LNT, you don't also get to say "and now we'll assume the function is even more favorable to my political agenda". Sure, it's possible the actual function makes it better for nuclear advocates. But it's also possible it makes it worse. Are the policy makers going to be on your side just because? I actually do get to say that. There is no evidence that insignificant doses of radiation do any damage. In the absence of evidence of harm after 40 years of study, we can safely assume that the harm is undetectably small and can be ignored. People can wave around numbers from a model, but the model is stupid and there is no reason to believe it. In the absence of a good model, I get to assume that we should make decisions based on the observed evidence. > OF course it's making a sensible point. As far as I've been able to determine, your mathematical argument is the gradient of a non-linear function is not constant. There is room for improvement in your explanation; unless that is your point in which case it is not a sensible point. The argument is that we should abandon the LNT model in favour of empirical evidence, ie, no harm done.
- deleted 7y ago