|
Authored by: Anonymous on Friday, November 02 2012 @ 02:36 PM EDT |
In practice, scientific fraud is a bad, but not terribly
significant issue. The vast majority of papers aren't
fraudulent. Albeit mechanisms for reporting scientific
fraud are not well-established. The current mechanism
appears to be...report suspected fraud to your dean and wait
for the dean to retaliate against you. (Based on NIH
whistle-blower study)
Publication of false-positive results based on selection
bias is a horrible issue that results for, eg, the vast
majority of cancer drugs can not be reproduced. That's a
disaster.
http://online.wsj.com/article/SB1000142405297020376480457705
9841672541590.html
If one hundred researchers test whether or not listening to
Mozart causes cancer...how many publications with P<0.05 are
expected? 5. How many negative results are published? 0.
How many researchers will do multiple tests until they get
P<0.5? 30.
The end result is that the majority of the scientific
literature in many fields is simply wrong in its
conclusions. Nature and Science are particularly known for
this issue.
While the punitive approach is not popular, I personally
favor long-term verification of the validity of the
conclusions on an academic's publications.
Conclusions that are well-supported by later literature
would boost one's score.
Conclusions that are later shown to be incorrect would
significantly adversely affect one's score.
--Erwin[ Reply to This | Parent | # ]
|
|
|
|
|