Bayesian Imposters
Sometimes, in conversation, people will ask “What do you do?” and they really just expect a one or two word answer. I struggle with this question. ‘Lecturer’ implies that most of my days are spent standing up in front of students, which is simply not true. If I say I’m a ‘researcher’, we inevitably have to follow up with ‘what kind of research?’ and this leads me to explaining that I do a bit of my own research on lots of topics, but often I just help other people with what kind of research they want to do. Sometimes I just say “I’m kind of a statistician”, which feels very wrong, but . . . I do end up correctly describing the majority of my teaching and collaborations. ‘Kind of’ is doing a lot of work in this answer, I don’t have a statistics degree, I need to look up everything I teach, and my expertise lies more in methodology, but a large volume of my working hours is spent wrangling data and running statistics.
I want to look a little bit at that discomfort today, in light of a recent paper that’s been preprinted: Silent Voices: Uncovering Womens’ Absence in Veterinary Surgery Publications. I ran the stats for this paper, which explores the discrepancy in the gender split of authors in the Veterinary Surgery journal. The short and to-the-point result is that there are fewer women publishing in Veterinary Surgery than we would expect, given the high proportion of women in the profession, and they rarely publish in senior authorship positions.
Being responsible for the analysis of this paper, and a collection of others on the same theme, felt tremendously stressful. I knew I wanted to take a Bayesian approach for a number of reasons, there were a lot of inherent assumptions in the data, I wanted to focus more on effect sizes that arbitrary cut offs for significance and interest, and I really like how easystats and bayestestr have made this type of analysis so much more accessible for people like me. But I had never formally been taught this.
I have been playing about with Bayesian analyses for a while now, running them by colleagues who do know them (and not had any gasps of horror regarding my approach), and even sneaked them in to a few papers, admittedly in a slightly under the radar way. I find myself really enjoying the approach and appreciating the way it really makes me think about the substance of my claims. The SEXIT framework helps with reporting, and working on this project definitely honed my narrative explanations.
But let me tell you, the nerves I felt when I posted those final analyses to our OSF repository and we submitted these papers. Who am I to make such bold claims about gender biases? What if I was leading my lovely colleagues up the garden path with my half-baked and poorly informed analyses? In our project team, we have often joked “act with the confidence of a mediocre privileged white man”, but I also want to highlight that there is value in those characteristics we perceive as ‘more feminine’. Given the replication crisis in science, maybe we should all act a bit more with the confidence of a mediocre privileged white woman like me.
So what does a mediocre woman do? Well, one of my big worries was about letting down the wonderful team who had invited me to work with them. when I started to worry about that, I tried hard to turn it around in my head to feeling grateful that this team of awesome colleagues wanted my help. I really had to work on this when they would ask me questions, particularly trying to remind myself that a question wasn’t to catch me out, it was to try and develop further understanding. Trying to approach these challenges with an open perspective rather than looking for the trap I think is a big step in trying to overcome imposter syndrome.
I also tried to have the courage of my convictions. Making all our code etc. available on OSF is part of what I think is important to improve our research process, but it definitely exposes us me . . . us to risk in terms of someone finding errors in what I’ve done. And I have had to take a deep breath, hold my hand up, and say “Good! Please do find the mistakes” because ultimately, I do want to provide the best version of analysis for this important topic.
This really reminds me of some of Madeleine Pownall’s work, particularly Navigating Open Science as Early Career Feminist Researchers, because we do have to acknowledge how intersectionalism applies in this case. One of the ways I am able to deal with imposter syndrome is being able to have enough privilege and security in my life to run the risk of being right. Maybe I am an imposter. Maybe I’m really shit at this. But I think I’m protected enough to take that risk. Not everyone has that luxury, male and female alike.
So next time you feel some imposter syndrome, I hope you know you’re not alone - and maybe we can start finding some positives in it too. But we definitely need to acknowledge the cost of open science practices and the price of trying to improve.