Causal Diagrams: Pitfalls and Tips

Abstract

Graphical models are useful tools in causal inference, and causal directed acyclic graphs (DAGs) are used extensively to determine the variables for which it is sufficient to control for confounding to estimate causal effects. We discuss the following ten pitfalls and tips that are easily overlooked when using DAGs: 1) Each node on DAGs corresponds to a random variable and not its realized values; 2) The presence or absence of arrows in DAGs corresponds to the presence or absence of individual causal effect in the population; 3) "Non-manipulable" variables and their arrows should be drawn with care; 4) It is preferable to draw DAGs for the total population, rather than for the exposed or unexposed groups; 5) DAGs are primarily useful to examine the presence of confounding in distribution in the notion of confounding in expectation; 6) Although DAGs provide qualitative differences of causal structures, they cannot describe details of how to adjust for confounding; 7) DAGs can be used to illustrate the consequences of matching and the appropriate handling of matched variables in cohort and case-control studies; 8) When explicitly accounting for temporal order in DAGs, it is necessary to use separate nodes for each timing; 9) In certain cases, DAGs with signed edges can be used in drawing conclusions about the direction of bias; and 10) DAGs can be (and should be) used to describe not only confounding bias but also other forms of bias. We also discuss recent developments of graphical models and their future directions.

Understanding Marginal Structural Models for Time-Varying Exposures: Pitfalls and Tips

“…That is why we do not start by drawing causal diagrams and use them only complimentarily in our illustration, despite the fact that they are indeed useful tools for explicating our assumptions in real data analysis. 35 The following "tips" emanate from two introductory subsections regarding the effects of point exposures and time-varying exposures. Then, we step into the main contents to understand the unique role of and distinction between inverse probability weighting, marginal structural models, and regression=exposure probability models.…”

“…That is, Figure 1 is one of the examples of causal diagrams that are compatible with our example data, where the stronger causal assumptions are implicitly imposed on. As we have documented earlier, 35 causal diagrams (when tied with underlying causal models) often represent the "finer" description of causal assumptions than counterfactual notation.…”

“…That is why we do not start by drawing causal diagrams and use them only complimentarily in our illustration, despite the fact that they are indeed useful tools for explicating our assumptions in real data analysis. 35 …”

“…[22][23][24][25][26][27][28] More importantly, a counterfactual approach has spurred extensive discussion on the assumptions for inferring causality from data and the conditions for specific statistical methods to work using, for example, causal diagrams. [2][3][4]6,[29][30][31][32][33][34][35] Yet, the most striking illustration brought about by the counterfactual approach may be that it can offer an elegant solution to the controversy surrounding the definition and estimability of the effects of exposures that vary over time. For example, initiated antiretroviral therapy (exposure) for acquired immunodeficiency syndrome may be intermitted after looking at the symptoms of pneumonia, which is a predictor of clinical outcomes (eg, death) but affected by the prior exposure, and thus considered as a part of the exposure's effects.…”

Pitfalls and Tips for Statistical Methods in Epidemiology: A New Series of Special Articles Has Started

“…1) Causal diagrams (ie, DAGs) (in this issue) 2) Marginal structural model 3) Propensity score methods (matching and inverse probability weighting) 4) Mediation analysis 5) Machine learning For the first special article in this theme, Suzuki et al summarized pitfalls and tips for causal diagrams, which are known as DAGs. 1 DAGs were first introduced by Greenland, Pearl, and Robins in 1999, 2 and they have been widely used by epidemiologist to select covariates. DAGs are a useful tool for summarizing the association between various variables, and numerous educational review articles have already been published.…”