This will be somewhat shorter summary of the second day of DSC 2014. There was a wider variety of presentations today, but I am only going to discuss those that touched on the main theme of the day, which was reproducibility.
Subscribers to the R-devel mailing list will recall a long discussion started in March by Jeroen Ooms called A case for freezing CRAN. Jeroen described the difficulties reproducing previous analyses (either those done by other people or done by oneself at a previous time) due to changes introduced in R packages. Jeroem proposed that a frozen version of CRAN be made available for each R release so that all users with the same version of R would have access to the same set of packages. Several presentations at DSC directly addressed this challenge
Simon Urbanek (AT&T) presented RCloud, an environment for collaboratively sharing R scripts. Data and code are pushed into the cloud so that the user does not care where it is stored. All user work is contained in a “notebook” which is automatically under version control. Notebooks can be shared by simply sharing a URL with another RCloud user.
JJ Allaire presented packrat, a dependency management system for R packages that essentially creates a separate library for each project. Packrat creates isolated, portable and reproducible environments for R packages, and does not require any changes to CRAN.
Michael Lawrence presented GRAN, a package management system, but my notes don’t allow me to say much about it here and the slides are not on the web site to jog my memory.
David Smith from Revolution Analytics presented the Reproducible R Toolkit (RRT) and its server-side counterpart the Marmoset R Archive Network (MRAN). This is the solution that most closely resembles Jeroem’s frozen CRAN. David suggested that it would be more useful to ordinary R users, as opposed to R developers who may prefer the approach taken by packrat or GRAN.
Special mention must go to Romain Francois, who drove through the night from the Rencontres R conference in Monpellier to Brixen to give a talk about Rcpp11, then went straight on to Paris to fly out to the UseR! conference in California. I told Romain he was crazy.
Overall this was an excellent meeting. It was very important to get together people who had been communicating by email only for a long period and I was impressed at the increasing investment in R by companies.