On 21 January 2021, the Environmental Statistics Section hosted a webinar on ‘causal inference in epidemiology and environmental health’. The event welcomed two emanant speakers in the field of causal inference, whose work covers applications in environmental health. This was a very topical meeting, which this was evidenced in the great turnout at the event. With approximately 115 attendees, the meeting aimed to engage both statisticians and practitioners in the methods and application of causal inference. Time was set aside for discussion with the speakers at the end of the meeting.
Firstly, Dr Marie-Abele Bind from the Harvard School of Public Health, delivered a very engaging technical talk on statistical approaches for causal inference. She focused on both her theoretical and applied work in this area. In particular, she referred to three main approaches to studying the impact of a treatment on a specific outcome of interest, namely Fisher, Neyman and Bayesian approaches. Marie-Abele focused on the first two of these approaches in her talk, explaining both analytical and simulation-based methods to detecting causal impacts based on a null hypothesis of no causal link. She also applied the methods to impacts of PM2.5 on daily deaths in Boston to provide evidence of a causal link between the two.
Secondly, Professor Neil Pearce from the London School of Hygiene and Tropical Medicine delivered a talk on controversies in causal inference, comparing his experience in causal inference in largely observational studies with those of randomised clinical trials. Professor Pearce’s experience here emphasised that despite RCTs being seen as the gold standard for determining causal effects, there have been many cases where causal studies of observational data have been instrumental in changing policy and scientific understanding. This is particularly important when RCTs are infeasible or do not work, such as in the impacts of smoking, or in studies prior to formal RCT theory being developed, such as in Hume’s 1748 ‘An Enquiry Concerning Human Understanding’. Professor Pearce’s overall conclusions were that we should be careful in rejecting studies that do not fit a single paradigm. In many scientific studies, a single ‘perfect’ study is not realistic and too narrow, and by considering all the evidence, we still have the opportunity to extract important causal links.
After the two talks, general questions were posed to the two speakers and there was detailed further discussion on the main themes of the two talks.
Dr Ben Swallow, lecturer in statistics at the School of Mathematics and Statistics, University of Glasgow