> It gets worse- From
> (snip)"Their report and *Science* article supposedly
> presented all the results of their exhaustive research.
> They did not. The authors fudged the data, and the
> "peer reviewers" and AAAS journal editors failed to
> spot the massive flaws.
Hooooold onto your horses, there.
The authors have already responded to these highly personal attacks being
made in the lay press rather than in the more respectable fora of online and
in-print journal comments.
Suffice to say, those doing the attacking are being told in no uncertain
terms that they need to find themselves a better statistician, hence the
title "Robust Statistics Explain Findings of Neonicotinoids Field
Experiment".
https://www.ceh.ac.uk/news-and-media/blogs/robust-statistics-explain-finding
s-neonicotinoids-field-experiment
http://tinyurl.com/yaeaq8n4
I'll let the article speak for itself.
I must comment on the "Mad Virologist" blog.
It was extremely entertaining, as I had never seen anyone attempt to
bludgeon a paper with Bonferroni corrections before. I should explain,
because it is actually funny in a math-geeky way. It is as if a pillow
fight broke out on the lectern at a conference.
When someone says "Statistically Significant", they mean that there is at
least a 95% confidence level that the results were NOT the result of chance.
While this is still a 1-in-20 chance that the results ARE by chance, one has
to draw the line somewhere, and for the history of modern science as we know
it, p=0.05 is the standard. (For the record, I bet on horses with only
20:1 odds all the time at Aqueduct and Belmont, and I do well enough to keep
at it.)
What Bonferroni does is say that one should toss all the data one has
collected from all the tests done into a single blender, and apply a test
for hypothesis significance. (So, in this case, they complain that data
from yards in Germany should be combined with data from yards in the UK,
rather than simply making individual statements about each country.) But,
here's the rub - the more data you toss into the blender, the more unrelated
individual tests you are aggregating, and the less your chance of finding
ANY significant result.
So, when someone says that you did not use "Sequential Bonferroni
Corrections", what they are really doing is saying that, for example,
sickle-cell anemia does not exist at all in humans, because it only
significantly appears in tests run in Africa. If you looked at the world
population of black people as a whole, the effect would no longer seem
significant.
This is such a lead balloon of an attack, it will prompt laughter among the
vanishingly small number of people who know what a Bonferroni Correction is,
and which end to point at the data to avoid shooting oneself in the foot.
***********************************************
The BEE-L mailing list is powered by L-Soft's renowned
LISTSERV(R) list management software. For more information, go to:
http://www.lsoft.com/LISTSERV-powered.html
|
