An Overview of Causal Directed Acyclic Graphs for Substance Abuse Researchers
Author(s): Michael Lewis and Alexis Kuerbis
Background. Within substance abuse research, quantitative methodologists tend to view randomized controlled trials (RCTs) as the “gold standard” for estimating causal effects, in part due to experimental manipulation and random assignment. Such methods are not always possible due to ethical and other reasons. Causal directed acyclic graphs (causal DAGs) are mathematical tools for (1) precisely stating researchers’ causal assumptions and (2) providing guidance regarding the specification of statistical models for causal inference with nonexperimental data (such as epidemiological data). Purpose. This manuscript describes causal DAGs and illustrates their use in regards to a long standing theory within the field of substance use: the gateway hypothesis. Design. Data from the 2013 National Survey of Drug Use and Health are utilized to illustrate the application of causal DAGs in model specification. Then using the model specification constructed via causal DAGs, logistic regression models are used to generate odds ratios of the likelihood of trying heroin, given that one has tried alcohol, marijuana, and/or tobacco. Conclusion. Granting the assumptions encoded in specific causal DAGs, researchers, even in the absence of RCTs, can identify and estimate causal effects of interest.