There are lies dam lies and statistics.
Statistics are used to give an indication of the likelihood (probability) of finding a particular result by chance. The smaller the P value p=0.01 is a 1 in 100 chance and P=0.001 is a one in a thousand chance.
We know that people doing basic science and publishing in Nature journal on experiments relating to MS have a problem grasping statistics as they often don't use the appropriate tests. (Is this arrogant?...Howeverl the data speaks for itself. Check out the link below on how the interesting finding reported in a Nature Paper becomes rubbish and an experiment that fails and needs repeating)
This will get my altmetrics up:-)
You may be interested in this post from Nature about the risks of interpretation of the P value.
I say... use the smack you in the eye test...If the data looks like a pile of pants it probably is.
Clinical trials generally have much better statistics because a trained statistician works out the results
However what does someone think about this article?
Here's a hihacked email that we got on this article (the person from beyond our shores has agreed to it's use-Thanks we had a laugh).
"I always find it ironic that Nature - the home of exaggerated medical junk
- has these types of articles. A few paragraphs caught my attention - as they are pathetic or naive
indicating Nature can't even discuss this topic without introducing politics and
journalism.
"p-hacking" is delusional cheating - nothing more or less. To perpetuate a
new name for it is sad.
Use of "plausible hypotheses" means we can't discover new things - the
earth remains flat. not a solution.
"size effect" note the nonsense about online dating was published in
Nature....
The problem is we have 3rd (rate?) scientists running the top [sic] journals and
politicians running our universities and funding agencies.
The solution for the issues in the article are simple. Ensure biologists
can count before leaving school, include independent replication in the primary
publication and relying on permutation methods to estimate an FDR (false discovery rate) and quote
'size effects' [or equiv] not just significance. That means, for example, at
least 2 background strains of TG (transgenic) mouse.....
I'd write to Nature, if I though anyone would listen - and explain people
play games with data because they need to or they have no job.... What the
liberal morons at Nature are trying to do, is excuse themselves from any sort of
rigour at all, creating a "p values useless" vacuum so they have greater
flexibility to publish more headline grabbing junk.
So someone may have a shock if they follow the altmetric links and come here:-)