Dennis Ogbe, whom I’ve written about before, has a nice example of why reproducible research matters. If you’ve read those previous posts you know that Ogbe is a PhD student in Electrical Engineering at Purdue. As such he reads papers in his field both for research and for his classes. One such paper had a sign error (possibly a typo) in the derivation of a set of equations used to process the data in the paper.
In a sense, this was small potatoes but Ogbe was trying to learn from the paper and tried to reproduce the simulation described in it. His results were wrong and he couldn’t understand why. It was only after checking the derivation by hand that he discovered the error and was able to reproduce the paper’s numerical results.
You can read Ogbe’s post for the details but none of this would have been necessary if the paper had included the actual code used to produce the results. Furthermore, if the paper had included that code it is more likely that the author would have discovered the error and corrected it before publication.
In a nice coda to the post, Ogbe demonstrates reproducible research in action by showing how the code used to build his table of results was included and executed in the post itself. Emacs and Org mode make that easy, of course, which is why the combination is so useful for publishing research in a reproducible manner.