ABSTRACT
Statistical models—including all statistical network models discussed in this book—describe probabilistic relationships between variables. They can tell us that if X is high, Y will likely also be high, but they cannot tell us what would happen to Y if we were to increase the value of X. In order to model the outcome of interventions, we need to go beyond statistical inference towards causal inference. This chapter introduces the reader to the core tenets and assumptions of causal inference from observational data. The reader will learn how to use directed acyclic graphs (DAGs) to express causal relations; what type of causal relations between variables imply which type of probabilistic relations and how to assess this using d-separation; how to assess, in a principled manner, whether a causal effect is confounded; how to compute causal effects using structural causal models (SCMs); and how to (possibly) discover causal relations from data.
