@incollection{ForaitaSpallekZeeb, author = {Foraita, Ronja and Spallek, Jacob and Zeeb, Hajo}, title = {Causal directed acyclic graphs}, series = {Handbook of Epidemiology}, booktitle = {Handbook of Epidemiology}, editor = {Ahrens, Wolfgang and Pigeot, Iris}, edition = {2. Auflage}, publisher = {Springer}, address = {New York}, isbn = {978-1-4614-6625-3}, doi = {10.1007/978-1-4614-6625-3_65-1}, pages = {27}, abstract = {In the past decade, causal directed acyclic graphs (DAGs), also called causal diagrams, have been increasingly used to determine confounding variables that must be controlled in order to estimate an unbiased effect between an exposure, an intervention, or a treatment on an outcome. We explain the reasoning of causal DAGs and why a DAG can be used to identify a set of sufficient minimal adjustment variables. In addition, we show how such a causal DAG can be drawn based on prior knowledge. We will discuss further approaches to determine sufficient adjustment sets. If prior knowledge is rare, unclear, or questionable, causal discovery algorithms can be applied to estimate a causal DAG based on data. These algorithms improved greatly over the last years. Nowadays, they can be easily applied to different data structures to estimate at least a partially directed DAG. We show the differences between constraint-based and score-based search algorithms and highlight the popular PC algorithm.}, language = {de} }