Author Interview: A brief introduction to mixed effects modelling and multi-model inference in ecology
PeerJ talks to Xavier A Harrison about the recently published PeerJ Life & Environment article A brief introduction to mixed effects modelling and multi-model inference in ecology.
Can you tell us a bit about yourself?
I’m a molecular ecologist at the University of Exeter, based at the Cornwall campus in the UK. Most of my research looks at how host-associated microbial communities can be beneficial for animal health. But I often get distracted by side projects looking at how we can best apply complex statistical tools to even-more-complex ecological datasets. I haven’t yet dared call myself a statistical ecologist because that sounds formal and triggers my impostor syndrome. Outside of work I spend my time paddleboarding and going on long walks along the Cornish coastal path with my dog, Leo.
Can you briefly explain the research you published in PeerJ?
Our paper arose from a series of stats workshops that Dr Rich Inger (University of Exeter) and I had been running over a few years at both the University of Exeter and Zoological Society of London, where I held a fellowship. These workshops covered a whole range of topics from understanding fixed versus random effects to the pros and cons of different model selection methods. The theme that emerged from all of these sessions was that whilst biologists often knew very well what statistical tools were available, it wasn’t always clear when and where one tool should be used over another, nor where some approaches may be less optimal depending on data and research question.
We thought there might be a gap in the literature that covered all these topics and issues under one roof. We wanted the paper to act as a roadmap for anyone starting out in of statistical modelling, and to highlight best practice in the fitting and interpretation of General Linear (Mixed) Models (GLMMs). Crucially we also wanted to maximise the opportunity for workshop participants to be involved, and so gathered an enthusiastic set of early career-researchers to lead dedicated sections on statistical topics they were passionate about.
Crafting a best practice guide to mixed effects modelling turned out to be tricker than planned! Our aim was to cover the vast range of topics and approaches surrounding GLMMs in biology, whilst not going in to so much detail as to overwhelm the reader, but also making sure to signpost where they could look if they wanted to delve into the specifics and more advanced topics. This was a tall order. But we are proud of the end product and hope it has been a useful guide for all manner of scientists.
What has been the impact for you and your field?
Our hope with this paper was that it would help people navigate the often overwhelming suite of tools available to biologists to model complex datasets. We also wanted to make sure it could cater to both novice users of GLMMs and also more advanced practitioners interested in the details of particular tools. And finally we wanted this paper to reach as wide an audience as possible, far beyond the core academic audience. We think we achieved this aim, partly because this paper gets cited so often, but importantly because it gets cited across so many different fields beyond just ecology and evolution.
I regularly get emails from scientists from many different sectors and backgrounds asking for advice on statistical modelling, or simply to say how useful they found the paper. I cant explain how great a feeling that is. Knowing the paper is being used and making an impact on the way people approach their analysis, and the interpretation of their data, is one of the best parts of the job.
How did you first hear about PeerJ, and what persuaded you to submit to us?
PeerJ was recommended to me by a colleague, who believed all research should be open access. They highlighted PeerJ as building a strong reputation in the OA space and couldn’t speak more highly of them.
Would you submit again, and would you recommend that your colleagues submit?
I’ve published several papers at PeerJ. In fact it’s the place I always send my statistical ecology papers because they have a proven track record of reaching a wide audience beyond just academia – like people working at NGOs who need to analyse tricky datasets. I have a few fundamental ecology papers that found a home at PeerJ too. The peer review process has always been relatively speedy and exceptionally fair and constructive, which is all you could ask for.
Anything else you would like to add?
Just to say thanks for the opportunity to talk about the work, and thanks to all the co-authors for making the paper possible!