Being a statistician in biosciences

2nd December 2022

Dr George Savva is our Statistician at the Quadram Institute. We spoke to George to find out about the importance of statistics in bioscience, his career and how he supports scientists.

“I support researchers at the Quadram Institute with the statistical aspects of their work, wherever this is needed.  This can mean contributing to study designs, evaluating statistical methods, coding and writing up statistical analyses or just talking to scientists about statistical approaches to their experiments or in what they are reading.

A mathematical artistic image which is pinks, red and white in colour, with swirling forms, including faint circular lines emanating from the centre.

In his spare time George makes mathematical art, partly inspired by his biological and scientific work

Variety and uncertainty

I’m a member of the Core Science Resources group, and we work with just about everybody else at the Quadram Institute. I’ve worked with most of the research groups in one way or another, and I get involved with science communication, audits, ethics, and with our commercial collaborators.

The famous statistician John Tukey (inventor the box plot, my favourite graph!) said that the best thing about being a statistician is that you get to “play in everyone’s backyard”.  It’s a great career for somebody who is interested in lots of areas of life science, from molecular to ecological to behavioural and social, and who enjoys science communication.  There is some maths and a lot of computer programming, but also philosophical questions about what we can or can’t say about the world based on the imperfect snapshot that our experiments and datasets provide.

Presenting data clearly and honestly, reflecting both our findings and the remaining uncertainty is an interesting challenge in almost every case.  We don’t want to miss potentially important results, equally we don’t want to claim to have found something that isn’t true.

Of course, nobody likes uncertainty, particularly after an experiment has finished. How to reconcile the need to present definitive ‘actionable’ research findings with the inevitable “yes, but…” is one of our most important challenges, particularly when we try to translate our smaller exploratory projects.

Pursuing a career path in statistics

My background is in maths. After studying maths and stats I completed my PhD in Bioinformatics at the John Innes Centre in 2006, and then spent the next ten years in medical and social statistics.  I worked in London, Cambridge, and Dublin before moving to Norwich in 2013 to lead a research group in cognitive epidemiology at UEA.

While leading a research programme was rewarding, I found that I enjoy teaching statistics and supporting others with their research more than pursuing my own. I also wanted to learn more statistics, particularly in the context of experimental sciences, to improve my maths and computing knowledge and to work on a wider range of science. My role as the institute Statistician at the Quadram Institute allows me to do exactly that.

Understanding all the different types of scientific work that we do at the Quadram Institute and the potentially enormous range of experimental designs and statistical and computational methods that go with that can be a challenge.  This is particularly true in emerging areas of work such as meta-genomics, where statistical approaches are less well developed, and we need time to evaluate and adapt statistical methods to the new types of data that our scientists are generating.

Supporting science projects from the start

A statistician colleague once described us as ‘research counsellors’ which is a good way to describe an important part of the role, when our input goes beyond the purely ‘statistical’.

Statisticians can be an independent critical eye on a project and help researchers refine their research ideas at the earliest stages.  Since we are ultimately responsible for data analysis and interpretation, we need to understand exactly what questions are being asked in a research study, why those questions are interesting and to whom, and exactly how we will address them through our experiments.  Drawing this out through conversations with those leading and running projects can help to refine many other aspects of the work.

I most enjoy my work when I can make a positive difference, either by improving a study design, suggesting an analytical approach, helping somebody to understand a statistical idea or to become more confident with statistics.  Contributing to impactful papers and getting new work started through successful funding applications is also extremely rewarding, with statistical inputs but also in critical appraisal, writing and through my experience on grant funding panels.

Many of the best statisticians have moved into statistics from other sciences.  Statistics isn’t necessarily about knowing a lot of mathematics but is about being able to think about science in a numerical way, thinking about how experimental conditions might bias results or add unwanted noise, and finding ways to account for these.

There are lots of accessible resources if you want to improve your statistics. I would recommend any scientist to make regular use of these to develop your understanding and practice rather than thinking of it as something you can’t do.

Statistics is the link between our data and the conclusions we can draw from it. High quality experimental designs and analyses are central to producing valuable science that funders and the public can have confidence in.

New approaches to data science via machine learning and artificial intelligence are exciting, but ultimately rest upon many of the same ideas, and so the skill set of statistical thinking, numeracy, computation, and communication will always be in demand.”

Related People

Related Research Groups