"John S. J. Anderson" wrote:
> >> This is like reading a yeast genetics paper without having the
> >> genotypes of the strains given, or a biochemistry paper without
> >> details of the experimental conditions.
>> Richard> Good point. This statement deserves wide dissemination.
>> Richard> Write a letter to Nature. Seriously.
>> Well, _someone_ should probably write such a letter. I'm not sure if
> I'm that person.
You sort of started this thread, why not write a draft and circulate it?
The response I've seen here so far does look encouraging.
> In the meantime, I'm going to have to devote some thought to the
> issues raised by alrichards at my-deja.com in a separate branch of this
> thread; my gut says he's wrong, but I need to get it to explain why to
> my brain.
Yep, he did raise one interesting point: code reuse may well result in
propagation of errors. Validation of code could be done similar to
clean-room reimplementation of software, if it's only about programming
errors. Won't catch more fundamental errors in the algorithm or the
model though (and that's what I'd be more worried about), and the level
of detail found in papers probably leaves too much room for ambiguity on
the side of the algorithm.
Michael
