2 November 2011

Reproducibility in Research

This post by Mario Pineda-Krch complains about the woeful lack of reproducibility in computational sciences.

This reminded me of Jake Bowers's good piece in the Political Methodologist from earlier this year about how to do reproducible computational political science. The article actually inspired me to completely switch over all of my new writing to Sweave. Sweave allows you to combine your R code and LaTeX documents. If you make your Sweave document and data available to readers they can completely reproduce everything in your article: the models, the table, the graphs, everything. 

RStudio makes using Sweave really easy (though I still use a text editor for writing much of the code since RStudio doesn't do spellcheck). 

Political economy and political science journals don't seem to have been keeping up with these developments. In fact, poli sci journals often require MS Word documents and don't allow you to submit Sweave documents. Few journal submission systems even allow authors to submit data and code appendixes before the paper has been accepted or R&R-ed.

No comments: