Posted by
Andrews on Thursday, July 24, 2008 9:28:01 PM
I wasn't going to cover this as Best of the Web did such a good job mocking it, but I saw the local news covering this as if it were serious, so I have to mention it. It appears a cancer researcher does not quite grasp the scientific process. Based on some unspecified preliminary results he is going against all of the accumulated research to date and saying that cell phones cause cancer. He is not deterred by the fact that no research supports this, in fact all points to the exact opposite conclusion. Nor is he troubled by the fact that his "preliminary data" has not been properly reviewed by his peers.
Basically, this "finding" has about as much validity as if I made a pronouncement about the causes of cancer. Yes, he is a researcher and knows quite a bit about medicine, but that does make him proof against mistakes. That is why we have peer review. Basically he is saying "I cannot make a mistake, listen to me."
And, because he is critical of technology, our Luddite media is saying "Sure, we will".
POSTSCRIPT
With the ubiquity of cell phones and the tendency for many to under report their actual usage, I have to wonder what his preliminary data looks like. Or is he using historical data to say incidence rose with cell phone usage. If the latter, I wonder how he controls for better diagnostic tools.
In short, I am pretty dubious that he could even establish this. But if he can, I still would not trust any "preliminary data" without a few other eyes reviewing his findings. I know from past experience in economic proofs that it is too easy to make the data say what you want, even when you do so unintentionally. It is easy to find what you want in a huge set of numbers. That is why review is as beneficial to the scientist as for the general public. It keeps them from deceiving themselves.
POSTSCRIPT II
One more thing, it is also possible that any single study may produce aberrant results. The larger the study, the less likely, but it still is possible. I recall an analysis we did of voting patterns by race, and the linear regression said that for every Indian in a given region, negative three people voted. In other words, not only did no Indians vote, but somehow they kept three other people from voting as well.
That was obviously nonsense, either we had peculiar data or we were using the wrong tool for our analysis, but it does show that sometimes it is possible, with the best of data and the best tools, to produce nonsense. Which is why you let other people review your work before making public pronouncements.
Unless, of course, you are seeking media attention, or are an extremist lunatic. Neither of which makes we put much faith in your findings.