I’ve only recently stumbled on structural equation modelling (SEM) and it’s potential uses for establishing causation in ecology. Amusingly, one of my best mates focussed her PhD on SEMs a in psychology, yet I never realised the potential of SEMs in my field. I stumbled on them reading Pillar & Duarte’s (2010) excellent paper looking at phylogenetic community structure (check out the article here: http://onlinelibrary.wiley.com/doi/10.1111/j.1461-0248.2010.01456.x/abstract), and being a naive ecologist was immediately drawn to the elegant pathways offered by this paper. They used SEMs to clearly understand what mechanisms drive phylogenetic niche conservatism. I could see the potential of the approach so tested it on my data and was impressed – it was time to find out more.
What I hadn’t realised was the enormity of the debate in the order of the debate between Bayesian vs. frequentist stats. As always, Dynamic Ecology provides a brilliant commentary – the response to the article is perhaps even more useful. As for me, I’m still undecided. At the very least using SEMs to make predictions to then test experimentally seems like a reasonable thing to do at the very least.