[Marxism] [pen-l] Fwd: Is a Controversial Nuclear Plant to Blame for Soaring Thyroid Cancer Rates in New York? | Alternet

Jeff meisner at xs4all.nl
Wed Dec 6 12:26:06 MST 2017


On 2017-12-06 18:28, Louis Proyect via Marxism wrote:
> 
> The problem is with cancer, however. Trying to find an environmental
> smoking gun is virtually impossible. Cancer clusters are just one
> example.

Well the sad thing is that I probably agree totally with Louis in 
regards to the policy issues regarding nuclear power (more than I would 
with David) and other environmental dangers expressed in the remainder 
of his post, below. But that has nothing to do with my point: knowledge 
gained empirically (and confidence in that knowledge) can be obtained 
using the scientific method, not through simply compiling what Louis 
admits is "circumstantial evidence," and what you learn (or don't learn) 
is a separate matter from what action you propose in relation to that 
knowledge (or lack thereof). Although our practical concerns are the 
same, I'm afraid Louis doesn't appreciate the use of statistical 
inference in establishing scientific conclusions, which is the only way 
he could say that doing so is "virtually impossible."

Statistics is an exact science (well, actually mathematics) concerning 
inexact phenomena (which is almost everything in reality). And by 
"statistics" I don't mean numbers that are measured or thrown around 
(aka "raw statistics") but the science whereby empirical measurements of 
random phenomena can lead to (or fail to lead to!) conclusions 
concerning the underlying system. Yes, there are many ways that 
statistics, even without making any errors in the math, can be 
intentionally misused or innocently misapplied, resulting in invalid 
conclusions. That goes for anything. But statistics has a bad 
reputation, because rather than the science of statistical inference, a 
propagandist will throw out numbers to non-statisticians and expect them 
to come to conclusions in their minds matching the intentions of the 
propagandist. But that is not how statistics is used in scientific 
research.

People might imagine that scientists make measurements and then perform 
calculations based on those numbers. But that's only half the story. 
Often you'll spend as much or more time not on those numbers but on 
analyzing the errors in your measurements and the resulting strength of 
your conclusions. For instance, I may measure a 2 degree increase in 
temperature when I add X to a solution of Y. That might mean that the 
reaction between X and Y generates that much heat. But before I could 
say that, and certainly before I could publish it, I would have to ask 
how precise my thermometers are. It could be that my thermometers are so 
crumby that either of their readings has an expected (rms) error of 1 
degree. Then I could hardly reach any such conclusion; even if there had 
been no heat generated in the reaction there is an 8% a priori 
probability of measuring a temperature increase as large as 2 degrees 
just due to the measurement error of my thermometers. On the other hand, 
even if those were the best thermometers I could obtain, I could repeat 
the experiment 50 times (making 100 measurements using the crumby 
thermometers). If the average measured temperature increase found from 
them is 1.32 degrees then I could conclude with extremely high 
confidence (in other words, I would stake my life on it) that there is 
an actual temperature increase. I could go further and claim 95% 
confidence limits for the actual temperature increase being between 1.26 
and 1.37 degrees. Having a 95% confidence means that using this proper 
procedure I would have successfully bracketed the actual value with a 
probability of .95. These statistical conclusions are all exact numbers 
based on sound mathematics but concern an underlying process that itself 
is never known with absolute precision. That is totally different than 
me casually saying "I measured this big number and it really looks 
convincing, doesn't it...."

So I'm sorry, but you very much can determine that radiation exposure 
leads to cancer without ever knowing that any single case of cancer was 
due to radiation. From a mathematical point of view you could do that 
best with a  controlled experiment where 100 people are exposed to 
radiation and comparing them to a test group of 100 who weren't. Of 
course that experiment is unethical (but has been done with animals) so 
you have to collect epidemiological statistics and these are more 
subject to confounding factors. The cancer rate increased from 3.3% to 
4.1% after the nuclear reactor was installed. For a community of several 
thousand, that would be a statistically significant increase. But there 
was also a chemical plant opened during the same period. And the ability 
to detect cancer was improved. So you compare with a second community 
also affected by the chemical plant but further from the nuclear plant. 
But then someone points out that the affected community had offered a 
tax break to cancer patients who move there. So maybe you aggregate such 
results from 100 similar nuclear plants. In other words, you can take 
steps to avoid the confounding effects and increase the significance of 
your measurements, and eventually you do reach conclusions. Because of 
all the potential unknowns affecting epidemiological data compared to a 
controlled experiment, there might be more uncertainty regarding your 
conclusions. But there are plenty of such studies which have properly 
linked ionizing radiation to cancer rates. Unfortunately this wasn't one 
of them.

- Jeff




  Put yourself in the position of a breast cancer victim living
> near a nuclear power plant. You are shit out of luck.
> 
> Even if the data indicates higher than usual contamination, how can
> you establish it led to thyroid or breast cancer?
> 
> The same thing is true of pesticides and herbicides. How can you
> establish that Monsanto is guilty of jeopardizing our health by
> selling products that after leaching into the soil cause bladder
> cancer, was the case with Sandra Steingraber?
> 
> You can certainly prove that drinking a fifth of whiskey every day can
> cause cirrhosis of the liver but what kind of experiment can prove
> that living near Indian Point causes thyroid cancer? It is only
> "circumstantial evidence".
> 
> Meanwhile, the best course of action is to overthrow the capitalist
> system and build a society based mostly on alternative energy sources.
> Even if the evidence is only circumstantial, that's no reason to risk
> the health of the human race.







More information about the Marxism mailing list