Hadley and Winston talk about the future of interactive graphics

-->

I went to my first R users group meeting in Minnesota and was treated to talks by two R heavy weights. First Hadley Wickham gave a talk about what he views as the most exciting and dynamic area of R development (interactive graphics) and this was followed by a short talk by Winston Chang showing what he has been doing to develop interactive graphics a part of the Shiny environment. Below are some take home messages and a few links that I thought were particularly interesting.

Hadley Wickham: First the slides for Hadley’s talk are on his github page. Hadley started his talk off with a short demo showing how interactive graphics could allow you to learn about your data. He has a clever dataset of 3d data points and these show no apparent pattern when viewed along any two axes. However when viewed from the correct angle we discover that there is a very distinct spiral pattern in the data this provides a nice aha moment in his talk. Next Hadley discussed the history of interactive graphics in R splitting them into 3 categories 1) those written in lower level languages, 2) those hacking existing R plotting functions, and 3) and browser based solutions. Many of the packages that he showed are not very realistic for everyday use or for investing a lot of time since they are no longer being actively developed.

Perhaps the most interesting part of Hadley’s talk was his discussion and demonstration of ggvis this is Hadley’s package that he envisions eventually replacing ggplot2. His goal is to continue to work on this and perhaps sometime in 2016 have it to the point that anything you could do in ggplot2 you can do in ggvis. The upside of this is that if you are already comfortable with ggplot2 you will have no trouble transitioning into ggvis. Hadley is using the same type of grammer based approach to building up more complex graphs from simpler elements.

Winston Chang gave a much shorter talk but illustrated some of the new interactive developments in Shiny. Despite my use of Shiny I was actually completely unaware of this development. Winston has added some nice examples on the Shiny articles page (bottom of right column).

Other interesting links from the evening:

Html widgets for R: allows you to use a variety of javascript visualization libraries within R

Hadley’s new book Advanced R

This semester I am teaching an experimental design course. However, three of my learning goals for my students are more general than experimental design:

munge a big dataset into different formats
use R to visualize their data and explore possible relationships
create scripts in R that process data do statistical test and output final plots for manuscripts

I wanted to justify the importance of learning good coding, scripting, computational skills to my students so I decided to mine the Evoldir mailing list for the last month or so. I first downloaded all of the adds for postdoc positions that were posted from December 1, 2017 to January 15, 2018. There were a total of 86 adds. For each one of these adds, I first determined whether any computational skills were listed as desired/required for applicants. Next, I counted the occurrence of requests for several more specific skills like knowing specific languages.

What I found was that 74% of the adds listed some form of computational skill as desired or necessary in the applicant. The bioinformatics category included all adds with vague statements like "competitive applicants will have experience running bioinformatic analyses of..." The misc. languages category included versions of C, awk, and java.

I should note that in some ways my approach underestimates the importance of computational skills. For instance, several of the adds that listed no computational skills are for departmentally funded independent postdocs. These advertisements usually list no skill requirements despite the fact that many departments and selection committees will none the less rate these skills as important in reviewing an applicants plans or previous contributions. The asterisks do not indicate anything about the add. Rather, these are skills that my students should begin developing in my class this semester if they work hard.

So the answer is yes. Having computational skills will almost always increase your productivity and this is becoming more and more widely appreciated.

* On a sad note more adds listed knowing Perl as a desired trait than unix or modeling - ugh!

Add a comment

Coleopterists Corner

Using Coleoptera to understand evolution; and lots of R

View comments