Only by explicitly excluding data from January and February were Sherman and Mangano able to froth up their specious statistical scaremongering.This is not to say that the radiation from Fukushima is not dangerous (it is), nor that we shouldn’t closely monitor its potential to spread (we should). But picking only the data that suits your analysis isn’t science—it’s politics. Beware those who would confuse the latter with the former.
He also points out that they hand picked cities, Boise but not Tacoma, for instance. Even with cherry picked data, the statistical signifigance of the data is slight. They average a buncha of numbers to reach a conclusion. If I apply the poission distribution to the whole data set, the probablity of observing the actual number (accepting the authors' assertion about the average rate), I would expect to see that number of deaths per week (12.5 vs. 9.5) about 1.7% of the time, just by the operation of chance, without any intervening mechanism. If I choose enough statistics and cherry pick them to support my conclusions, I will have absolutely no problem finding outcomes that occur only once in 60 trials. What is significant is that they cannot find more convincing evidence, given the way they tortured the data.
Do you know of any good data monitoring source on Fukushima? The media is very quiet.