At 06:23 AM 7/6/01, Jeroen wrote:
>At 05:53 6-7-01 -0500, Adam Lipscomb wrote:
>
>> > > > That's because both sides have a different approach. Assume someone
>> > >reports
>> > > > a significant increase in the number of cancer cases near a nuclear
>>power
>> > > > plant. The pro-side will say "there is no evidence that it is caused
>>by
>> > >the
>> > > > power plant". The con-side however reasons: "radiation (like from a
>> > >nuclear
>> > > > plant) can cause cancer; if there aren't any other likely suspects
>>that
>> > > > might have caused this high number of cases, then logically it must
>>have
>> > > > been caused by the power plant".
>> > >
>> > >But then the pro side rebuts with, "Funny you should say that, since
>>we've
>> > >done a statistical study that indicates that there is not, in fact, a
>> > >statistically significant increase in cancer rates near this plant, and
>> > >here's the evidnece we gathered." The con-side then repeat, "But we've
>>got
>> > >these reported cases!" and they never offer any evidence beyond
>>anecdotes.
>> >
>> > And when the con-site then asks the pro-site "if it wasn't the power
>>plant,
>> > then wat *did* cause this outbreak", the answer is either "coincidence" or
>> > "er... er... er...".
>>
>>The pro side has offered a reply to that - it's in the data. The "cluster"
>>of cases you're reporting is not really there. The cancer rates are not
>>different from the rates elsewhere.
>
>If you re-read the above quotes, you'll notice something strange: a
>mysterious disappearance.
>
>The scenario I depicted was "assume someone reports a significant increase
>in the number of cancer cases near a nuclear power plant". The use of the
>word "reports" indicates use of factual information (like medical records
>from the local hospital). Despite those medical records, the pro-nuclear
>side says "the cluster of cases you're reporting is not really there".
>
>If medical records show a significant increase (say, three times the
>national average), how can those cancer cases suddenly have disappeared
>when a study is done to determine the cause? Magic?
>
>NOTE: this scenario is not linked to any specific nuclear plant; it's a
>scenario, a hypothetical situation where medical records show a large
>increase in cancer cases. The cause is unknown, but some people consider a
>nearby nuclear plant to be a likely suspect.
No. The cancer cases do not suddenly "disappear." They are real.
The question is whether or not the fact that those cancer cases are found
in the vicinity of a nuclear power plant is anything more than a
coincidence or if it suggests that there is something about the nuclear
power plant that causes an increase in the cancer rate in its vicinity.
As long as we are talking hypothetically, let's change the example a
bit. Suppose that _20/20_ or _60 Minutes_ or _Dateline NBC_ or some other
program of the sort runs a report that says that within one mile of the
Wormwood Nuclear Plant environmentalists have found only one third the
national average number of endangered spotted owls, and they have the facts
to back it up. That makes it sound like the plant has either driven away
or killed two thirds of the spotted owls that should be living
there. Before painting a sign reading "OWL KILLERS" and going out on the
protest line, though, I might ask myself a couple of questions. First,
what is the national average number of spotted owls living in an arbitrary
circle of one mile radius? Is the average closer to 3 or 3000? If the
latter, then 2000 owls are missing. If the former, 2 owls are missing from
that 3.14-square-mile area. In both cases, the area around the nuke plant
has "only one-third of the national average number of endangered spotted
owls." Which of those two cases seems more troubling, though?
Also, what is the _variance_ in the spotted owl population density? (Or
the "standard deviation," which is the square root of the
variance.) Assuming that on average there are a significant number of
spotted owls in a circle of one mile radius (i.e., that the average is more
like 3000 than 3), then if everywhere else in the country the number of
spotted owls in a circle of that size is never less than 2900 or greater
than 3100, the circle around the nuke plant stands out. If, however, the
variance in the rest of the country is large, with some such circles that
do not contain nuke plants having only a few hundred spotted owls while
other such circles have 5000 or 6000, then the population of 1000 (= "1/3
the national average" of 3000) in the circle around the nuke plant is not
at all out of line.
So, if (hypothetically) someone in the media reports that the rate of a
certain type of cancer among people within one mile of a particular nuclear
power plant is "three times the national average," as verified by records
from the local hospital, the report makes it sound bad, but I really don't
know much at all until I track down the study that the news report was
based on and find out the values of the mean (= "average") and variance of
the occurrence of that particular type of cancer among the population of
the nation as a whole. Armed with this additional information, I can then
make a reasonable judgement about whether the additional cases of cancer
within one mile of the nuclear plant really indicate a significant increase
over the average or are explainable by normal statistical variation about
the mean. In each of the real-life cases that have been reported upon so
far, any cases where the rate of cancer near a nuclear power plant was
higher than the national average have turned out to be well within the
expected range of the cancer rates based on statistical variation. Add to
that the fact that in a nuclear power plant of modern Western design the
amount of radiation that escapes the containment facility around the core
is much less than the normal background radiation, there is no evidence to
back up an assertion that such nuclear power plants contribute to
additional cases of cancer in their vicinity.
Also, as I suggested in an earlier post, what if (hypothetically) a study
(with medical records from the local hospital to back it up) found that the
rate of cancer was three times the national average within one mile of a
windmill farm? Would anyone assume that windmills cause cancer?
-- Ronn! :)
