China Third Country To Be Hit By 'Brown Tide'

There's a fascinating story with Greenpeace and GM corn. The folks making the GM corn did a study where they got both their GM corn and conventional corn that was as similar as possible, and then fed both to lab rats. They weighed the rats every week, then took the rats apart after a while and assayed... everything. Organ sizes, weights, chemistries, etc. They concluded that there were no significant differences.

Greenpeace sued to get the raw data, something I think they have a right to (since that study was used as the basis for approval). They got some folks (grad students in Germany, I think) to do their own statistics, which concluded that GM corn caused a statistically significant increase in growth rate for male rats and a statistically significant decrease for female rats. I looked at what they did, and it turns out they made a sophomoric statistics error that I teach, well, sophomore undergrads not to make.

What they did, essentially, was to neglect the fact that limited-sample-size uncertainties in "weight of rat at 6 weeks" and "weight of rat at 7 weeks" are correlated when they tested for statistical significance. Of course they're correlated -- they're the same damned rats! (In technical language, they calculated chi-squared based on the naive standard-errors-of-the-mean, rather than on the full covariance matrix which is required for [strongly] correlated data.)

If Greenpeace can't even get undergrad stats right in one of the cases where they *have* shown their work (and it's wrong) then I see no reason to give them any credibility unless someone who's better at this than they are checks their work.

Source: http://rss.slashdot.org/~r/Slashdot/slashdotScience/~3/FUxJsj0F14A/china-third-country-to-be-hit-by-brown-tide

chimpanzee the lucky one pittsburgh pirates mariners mets shades of grey pittsburgh penguins