> 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.
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.
That 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.
OF 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?
> 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.
> There is no evidence that insignificant doses of radiation do any damage.
A single high energy photon causes DNA damage which is easy to replicate. The theory you are supporting refers to rates of cancer, but has zero direct evidence from cancer rates to support it. And that’s the problem, science defaults to the older theory which in this case is the linear model.
> A single high energy photon causes DNA damage which is easy to replicate.
Which is relevant but unpersuasive; for me to agree +-1 photon to make any difference to getting cancer I'd have to suspend everything I know about statistics. We get hit by an ungodly number of high energy photons. If one cases cancer, there are probably others.
At some point, the doses become to small to matter.
> science defaults to the older theory
No it doesn't. Science defaults to the simplest/most likely theory and chooses the balance between those two things based on evidence. Age of the theory has nothing to do with it.
We have evidence that the flat earth theory is wrong, for example. Nobody can claim it is a serious contender despite being an ancient theory.
Similarly, there is evidence that the LNT is stupid. So while people can claim it is a contender because the evidence is weaker than for the globe it shouldn't be a default. The default should have a threshold below which we admit we've seen no evidence of any harm and people have been looking for decades. There is a lot of cancer out there naturally.
And even then the most likely truth of the matter is that nuclear accidents are good for cancer outcomes because it makes people actually screen for cancer. Then they catch all the non-nuclear related stuff.
It’s a default specifically when you have no evidence to replace the existing theory. Flat earth fails on evidence, radiation threshold fails for lack of evidence to support it over an older theory. It’s a very different situation.
It’s known that it takes multiple mutations to get cancer, but people are constantly getting cancer thus there must exist people that have almost gotten cancer and will get it from a single unlucky photon in a specific ___location. And other people who will get every other mutation eventually.
Thus from all known evidence we have the minimum threshold for extra radiation causing cancer and everything else being equal is exactly one photon. Calculating specific odds is much more difficult, but any theory that has some radiation level as absolutely safe is wrong.
That’s not to say it’s a pure linear relationship, but the slope can’t be zero.
Note, I am not saying a linear model is accurate and I doubt it is. But the threshold has both massive theoretical issues and zero direct support. Putting it next to string theory as an interesting theory, but completely untested science.
> Flat earth fails on evidence, radiation threshold fails for lack of evidence to support it over an older theory.
We've got oodles of evidence. People have been looking to black-eye the nuclear proponents for 40 years after Chernobyl and nobody has turned up any evidence. At some point, maybe 20 years, maybe 30 years or even 40 years the overwhelming lack of any evidence of harm becomes evidence that none was done.
> That’s not to say it’s a pure linear relationship, but the slope can’t be zero.
Well no that isn't true. Stepping out of the nuclear realm, this is very similar to arguing that a railgun will blast a hole in a building, a wrecking ball will knock a smaller hole in a building and therefore enough youths punching buildings will statistically knock one over.
That isn't going to happen. If that logic works out we are talking a seriously troubled building. A building that a stiff breeze could knock over, and a building that is almost surely knocked over before the youths get to it to punch it. The slope becomes indistinguishable from zero - in fact, it probably is zero. I doubt you'll want to contest that but if you do I would encourage some reflection on what a feeble hill that is to fight on vs real-life policies that actually matter for bringing energy to millions of humans and saving a measurable number of lives vs coal power or even solar installation.
This argument that thresholds are impossible is taking no cues from all the other forces, where there are clear thresholds below which no damage is done. An equivalent number of spaced out flicks deals nowhere near the damage of a punch, and there is a threshold below which force does no practical harm. Your argument we shouldn't take cues from the other forces is the first people to look at the problem drew a straight line through the data and therefore you don't want to accept that vanishingly small forces are probably irrelevant. An opinion held in an unscientific defiance of the preponderance of evidence, I cheerfully add.
There is zero direct evidence for departure from the LNT at the doses that would be relevant for the larger population in a nuclear accident. The reason is simple: for any individual, the extra chance of cancer would be so small that it could not be detected in the very large background of cancer from all sources.
Suppose 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.
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.