On 7-12-2009 4:42, Darren New wrote:
> Neeum Zawan wrote:
>>     But peer review doesn't demand repeatability. Well, OK - it should 
>> demand that the setup be described in a way that someone can repeat 
>> it, if that's what you mean.
> 
> That's what I meant, essentially.
> 
>>     Peer review can be a good first order filter, but let's not glorify. 
> 
> Definitely. I really think it's the repeatability that's *more* 
> important. Sure there are lots of papers where you can't repeat the 
> results, but I suspect the harder it is to repeat, the less fundamental 
> the result is.

There are some things next to peer review that help honest reporting of 
data: you need to keep records of what you have done. Even if the paper 
does not give enough information (been guilty of that myself, but never 
intentionally), for a number of papers, someone from the editorial board 
(or someone on behalf of it) can come visit your lab to check the data 
afterwards. For every figure in the paper you have to be able to find 
the experiments it was based on and the raw data. If you publish data 
based on a clinical trial you also have had that trial registered 
somewhere *before* the inclusion of the first patient, including a full 
and reviewed protocol. (A couple of colleagues were at a seminar 
explaining all the procedures recently. One of the more interesting 
facts was that that registration procedure will cost about 40 kilogram 
of paper per site were the trial runs (a number apparently based on real 
life experience). So clearly the procedure is not optimal yet, but at 
least it will help reducing fraudulent trial reports)

Post a reply to this message