Computer scientist Judea Pearl and colleagues have developed an approach for representing causal relationships referred to as causal directed acyclic graphs (DAGs). DAGs allow researchers to encode their beliefs about causal relationships in mathematical objects called graphs. Assuming the relationships in a given graph obtain, there are certain rules which researchers can use to determine whether the causal effect of interest can be identified. The answer to the identification question also provides guidance regarding which variable (or variables) need to be included in a statistical model to estimate a causal effect of interest.
Another use of DAGs is to determine if a researcher’s causal assumptions are supported by the data. Once an analyst has encoded those assumptions, certain conditional independence relationships between variables are implied. . If these relationships obtain in a given data set, then the DAG is consistent with those data, in the sense that the causal processes encoded in the DAG could have generated those data. If the implied conditional independencies aren’t actually found in the data set, then the causal processes encoded in the DAG couldn’t have generate those data.
Content. This presentation will provide an overview of these two uses of DAGs. I will describe the basic concepts of DAGS, including nodes, directed edges, paths, d-separation, and the backdoor criterion. The focus will be on how these concepts are related to specification of statistical models, intended to estimate presumed causal effects, as well as search for such effects. All of this will be done in relation to a data set on drug use. I’ll also offer critiques of DAGs and describe how they are different from but may be complementary to the Campbell and Rubin frameworks previously discussed.
Implications. Unlike many other quantitative innovations in social work and the social sciences, DAGs aren’t a new statistical technique, or set of techniques, for estimating relationships between variables. Rather, they are more similar to the Campbell and Rubin frameworks for causal inference, providing an internally-consistent set of guidelines for imbuing a statistical association with causal meaning. They are compatible with the major statistical methods typically used in social work research, including regression, structural equation modeling, and propensity score matching. The innovation DAGs provide is a way of thinking about as well as offering guidance regarding which relationships should be estimated in the first place.