Thoughts from the Australian Statistical Conference 2014

I started this post as a straight review of the Australian Statistical Conference 2014, but it turned into something else about large conferences and what makes me excited about statistics (spoiler alert, the answer is data science).

This year’s conference was in Sydney and was held jointly with the annual meeting of the Institute of Mathematical Statistics. Being held jointly with the IMS had two main effects:

  1. This was by far the biggest statistics conference I’ve been to, with some 500-600 delegates and up to 8 parallel session.
  2. There were many ‘mathematical statistics’ talks, which, I’ve come to realise, just aren’t of much interest to me.

The trouble with large conferences

Okay, I’m gonna start with the bad. Such a large program leads to well-known frustrations when there seem to be no talks of interest in one session and then three fascinating talks on all at the same time. It makes me wonder whether I’d really enjoy some of the even bigger conferences such as the Joint Statistical Meetings or American Society of Human Genetics meeting, both of which attract more than 6,000 delegates.

I favoured the keynote sessions and those sessions including people who I already knew I wanted to hear from, sessions including friends I wanted to support and sessions in my research area of bioinformatics.

I’ve also learnt that it’s quite alright to skip a session if nothing takes my fancy in order to have a chat with an interesting person (my definition of ‘networking’) or to sneak off to grab a decent coffee.

The upsides of large conferences

On the other hand, big conferences do attract some great speakers and interesting people to meet and chat with, it’s often just harder to find them. And, of course, the conversations during the breaks between sessions and afterwards are the best parts of a conference.

It was great fun to make friends with other young statisticians - hi, Hannah (@HannahLennon_) and Eve (@lattesipper) - and catch up with old friends, as well as get the opportunity to meet and chat with a couple of statisticians who I really admire, including Matthew Stephens (@mstephens99) and Hadley Wickham (@hadleywickham).

Matthew gave a really enjoyable IMS Medallion lecture, False Discovery Rates: A New Deal. What I particularly liked about Matthew’s talk is that he spent the first ten minutes discussing how to do good work, including organisational skills, git/GitHub and knitr. See here (pdf) for a previous version of this talk and GitHub for the ashR R package that implements the adaptive shrinking techniques Matthew described. Peter Donnelly’s talk in the same session on Statistical Challenges in Genomic Discovery was also fantastic.

Hadley spoke to a about the tidy-transform-visualise-model cycle of data analysis (pdf), including the tidyr, dplyr and ggvis R packages he develops. Also, Hadley is writing an incredibly helpful book on Advanced R Programming that can currently be accessed for free

What makes me excited about statistics?

The two talks I highlighted hint at my favourable bias towards people who work to tackle ‘real’ problems or who help other do the same. These problems are certainly not limited to academic or industrial research. In fact, most are probably tackled outside of these environments, which means I don’t hear so much about them since I work in an academic research institution.

For example, Sheila Bird’s talk on improving prisoners’ access to harm-reduction - such as Hepatitis B immunisation - which prompted a change in public policy, was a highlight of the conference. I’d love to see more talks like Sheila’s about statistics playing a role in society and public policy and presented so entertainingly.

I don’t mean to be pejorative when I describe these as ‘real’ problems - I don’t even have a good definition of what I mean by ‘real’ problems, except that “I know it when I see it”. Roughly, what I mean by this is that I like people who actually do data analysis or who help people analyse data, which includes people who write good software.

There’s a fair bit of hand-wringing these days at statistics meetings about whether statisticians have ‘missed the boat’ when it comes to data science and whether there is a difference between data science and statistics (yes, I think there is and that the difference is only getting bigger). While I missed his talk, Bob Rodriguez, a director at SAS and a former president of the American Statistical Association, reportedly had some astonishing numbers on how few people are graduating with a statistics degree compared to how many jobs are predicted to be created that require data science skills. On these numbers alone, it seems “statistics”, at least in its current form, doesn’t stand a chance.

All this leads me to conclude that what excites me about statistics is simply data science. Jan de Leeuw’s tweeted sentiments do well to capture my current feelings: