The R Packages of UseR! 2016

by Joseph Rickert

It is always a delight to discover a new and useful R package, and it is especially nice when the discovery comes with at context and testimonial to its effectiveness. It is also satisfying to be able to check in once in awhile and get an idea of what people think is hot, or current or trending in the R world. The schedule for the upcoming useR! conference at Stanford is a touchstone for both of these purposes. It lists 128 contributed talks over 3 days. Vetted for both content and quality by the program committee, these talks represent a snapshot of topics and research areas that are of current interest to R developers and researchers and also catalog the packages and tools that support the research. As best as I can determine, the abstracts for the talks explicitly call out 154 unique packages.

Some of the talks are all about introducing new R packages that have been very recently released or are still under development on Github. Others, mention the packages that represent the current suite of tools for a particular research area. Obviously, the abstracts do not list all of the packages employed by the researchers in their work, but presumably they do mention the packages that the speakers think are most important or in describing their work and attracting an audience.

The following table maps packages to talks. There are multiple lines for packages when they map to more than one talk and vice versa. For me, browsing a list like this is akin to perusing a colleague's book shelves; noting old favorites and discovering the occasional new gem.

I am reticent to try any conclusions about the importance of the packages to the talks from the abstracts alone before the talks are given. However, it is interesting to note that all of the packages mentioned more than twice are from the RStudio suite. Moreover, shiny and rmarkdown are apparently still new and shiny. Mentioning them in an abstract still conveys some cachet. I suspect that ggplot2 will figure in more analyses than both of these packages put together, but ggplot2 has become so much a part of that fabric of R that there is no premium in promoting it.

If you are looking forward to useR! 2016, whether it attend in person or to watch the videos afterwards, gaining some familiarity with the R packages highlighted in the contributed talks is sure to enhance your experience.

Comments

[David Ibarra]

Typo: lpSolveAPI and not lpSolveAPO.

Regards

[Tal Galili]

Hi Joe,
I also wanted to add heatmaply to the mix, but didn't get to release it before submitting the abstract (I will, however, speak about it, along side d3heatmap).

See you soon :)
Tal

[Berry Boessenkool]

Very nice post. Sadly I won't attend in person, but the videos will hopefully be awesome.

Did you do automated scraping of the abstracts to get the package mentions? Or has that all been manual?

[Joseph B Rickert]

This was mostly a manual process. I did write some simple R code to scrape the abstract description fields, break them into words etc. and then look to see if any of these words were in the list of CRAN packages. (The available.packages() function makes this easy) However, I soon realized that this approach is way to simple to solve the problem in general. Look at this first sentence from the abstract for the bamdit package:
In this work we present the R package bamdit, its name stands for "Bayesian meta-analysis of diagnostic test-data".

"its" is an R package on CRAN, but obviously not relevant to this talk. I need a way to establish the context to sort this out. Maybe for this relatively small corpus someone (probably not me) could write a few simple rules that would get 90% of the way there. But in general, I think this is a pretty difficult and interesting problem.

[Joseph B Rickert]

I have been advised that the package listed as WTAQ2" in the table above is actually called "kwb.wtaq" and is available under Github: https://github.com/KWB-R/kwb.wtaq.