Topic: directed graphs

Translate biological hypotheses into causal questions using potential outcomes and directed graphs. The practical programme identifies confounding, mediation and collider bias and implements matching, propensity scores, inverse weighting, g-computation and doubly robust estimation. It then develops causal machine-learning approaches, including forests, targeted learning, uncertainty quantification and conformal prediction, before applying mediation analysis to multi-omics integration and reproducible R/Python workflows. Sessions: 2–6 November, 14:00–18:00 daily, Europe/Berlin (CET, UTC+01:00)

The course teaches causal reasoning for ecological and evolutionary studies where controlled experiments may be impractical. Participants translate biological hypotheses into directed graphs, identify confounding and selection structures, and fit Bayesian structural equation models in R. Practical work covers posterior diagnostics, mediation, mixed response types, hidden causes, competing models and phylogenetic dependence. The programme emphasizes what observational evidence can and cannot establish, with opportunities to develop analyses for participants’ own systems. Online sessions run 14: