I am very much interested in the idea of electronic
publishing, not only of papers but also of the data sets upon which
the papers are based and of the code for the analysis of the data. I
do neurophysiology: multi-electrode recording. There have been
several recent and somewhat controversial articles of late and it
would do the whole neuroscience community some good if someone was to
set up a site where the papers, the data on which they were based, and
the analysis code used in the papers was made public. This would
ensure that not only the papers but the data in them could be reviewed
by others in the field. Also, having the code for a new analysis
procedure made public will allow others to apply the method to their
system and for any problems in the approach to be discovered.
We are considering such a thing in the Woodward lab at Bowman
Gray. Our lab uses matlab and stranger extensively for data analysis
and we will make our data sets and analysis procedures publically
available. Others with matlab and stranger can then look at the data
directly and try out our new analyses on their own data. Of course,
this won't happen until the stuff is published. In my case, my thesis
papers should go out sometime this summer and thereafter I plan to put
everything on our ftp site.
This idea is common in statistics, especially in the wavelet
community. I have been using some of these methods and have gotten to
know some folks who have argued for making research _reproducible_.
That is, after you publish an article, you put a data set and the code
needed to reproduce the figures in the paper on a web site so that
others can literally reproduce your research methods and results. The
folks at Stanford have really done a good thing with their papers and
through their analysis package, WaveLab. I think those of us who use
computer-intensive methods in the neurosciences could do our community
some good if we try this ourselves. This includes not only
neurophysiol types like myself, but also those working on cellular
physiology and computational modelling. I hope to bring this idea up
at the CNS meeting this summer.
What do you think?