Abstract: Biomedical data are characterised by large heterogeneity. Technical challenges result in noise and biases, e.g. through measurement errors, but more importantly, biological variability between individuals and conditions is large. This is even more the case in disease, which is a deviation from the normal healthy state, and each patient differs from this norm in a different way. Statistical modelling is concerned with delineating and understanding the different sources of heterogeneity. Bayesian statistics offers a unified approach to the modelling, which allows us to keep track of all sources of variability, even in complex hierarchical model setups. I will illustrate the potential of Bayesian modelling for complex biomedical data structures through examples from precision medicine and cancer drug screens.
Host: Rein Aasland