Mid-Semester R Psychology Workshop

Over the September mid-semester break, we worked with Maggie Webb, president of University of Melbourne’s Graduate Researchers in Psychological Sciences (GRIPS), to deliver a Software Carpentry workshop focused on Unix shell, version control with Git, and the R statistical programming language. In addition to people from GRIPS, we were also able to welcome researchers and students from the Graduate School of Education’s Centre for Positive Psychology (CPP).

In total, 35 people attended. Instructors included Alistair Walsh (shell), Tim Rice (Git), and Scott Ritchie (R-stat); the excellent Simon Lilburn was kind enough to attend as a helper.

This was the first workshop we’ve ever run with people from Psychology and it was a great success. The students were bright and enthusiastic; some already had strong computational backgrounds due to their need for intensive statistical calculations in their research. Others were learning everything from scratch and it was incredible to see how much they could pick up in just two days.

The structure of the workshop was to teach shell the first morning and Git the first afternoon, then spend all the second day on R.

Alistair delivered the shell lesson in highly entertaining style, and furthermore came up with the idea of collecting feedback on stickynotes after each set of lessons. This way we get more focused feedback about each part of the workshop, and before memories start to fade.

To teach Git, Tim used the traditional commandline-based lesson from Software Carpentry. Feedback indicated that people found it interesting, but a large proportion also said they felt left behind. While on the one hand we try to reassure people that they don’t have to keep up with everything the instructor is doing, on the other it’s understandable to want to immediately try everything out. Alistair and Tim have begun developing ways to make the first exposure to Git simpler for non-computational people, but we weren’t yet ready to trial these ideas in September.

On the second day, Scott provided his usual excellent delivery of the R lessons using RStudio. Scott has not only taught this many times before but also was the main author of the lesson material, resulting an a very polished presentation. Feedback on the R lessons provided the kind of bellcurve we usually aim for, with some people finding the material too slow and others finding it too fast. So long as we have even numbers at both ends, it was probably about the right level of difficulty.

Some people also expressed the wish to have more statistical applications rather than abstract programming concepts. This is understandable, but we’ve so far tried to avoid going down this route for a couple of reasons.

Firstly, we don’t want to tread on the toes of our worthy colleagues in the Statistical Consulting Centre (SCC). SCC provides statistical support and training to graduate students and researchers, in an analagous way to the computational support we offer in Research Platforms. They are much better equipped in terms of training and experience to impart statistical understanding, and to show how to augment a statistical analysis using various types of software, including R. They can consult one-on-one with current university researchers ― including graduate students doing research towards a thesis ― and several times per year run a Statistics for Research Workers short-course.

If Research Platforms attempted to provide statistical training within our R workshops we would probably just end up botching it because we’re not statisticians. We hope it will be clear that people who want help with statistics should go to the SCC, and people who want help with programming and computation will come to Research Platforms.

The second reason to abstain from teaching statistics in our R workshops is that in such a brief workshop, there is always going to be something that gets left out. We choose to leave out the parts that people can learn from other sources after they’ve got an initial conceptual foundation in place. If you’ve never programmed before, it’s a lot more important to leave our workshop knowing what a for-loop is than to know how to do an ANOVA or fit a generalised linear model in R. If anyone can think of a way to cram all the fundamental programming concepts and all the statistical applications into a two day workshop, please come and talk to us because we’d be rapt to know how to make it possible!

However, we have taken it on board that the non-statistical nature of our training should be emphasised more clearly up front. In future we can make sure to highlight the Statistical Consulting Centre as the place to go for training and support in data analysis. Some of our people in Research Platforms and broader community also offer Data Carpentry workshops, so one possibility is to offer more of these to audiences who might be less interested in computational concepts and more interested in the specifics of working with data.

A final innovation that we trialled for the first time at this workshop was to hand out Hacky Hour drink vouchers to all attendees. These were greeted with much enthusiasm and we’ve been happy to see an increase in numbers of people coming along to Hacky Hour after the workshop for follow-up conversations about their computational problems.

Overall, we all had a fantastic time and we really look forward to working with people in GRIPS and CPP again in 2016 and beyond!