Setup

Loading relevant package. Here we will be using ggdag and setting a theme that is provided by the package.

library(ggdag)
theme_set(theme_dag())

Build DAG

Build and plot your DAG using the “dagify” and “ggdag” function, respectively. Note that the variable that is on the left of the tilde is what the arrow is going into. Additionally, the “+” symbol is used to graph a collider.

NV_CVD_dag<- dagify(CVD ~ NV,
                    NV ~ income, 
                    income ~ occupation,
                    CVD ~ occupation,
                    PA ~ occupation,
                    PA ~ NV,
                    CVD ~ PA, 
                    CVD ~ income,

#physical activity is a common child of income and occupation
                    PA ~ income + occupation,

#create labels for variables ("\n" denotes a space)
       labels = c("CVD" = "Incident\n CVD", 
                  "NV" = "Neighborhood\n Violence",
                  "income" = "Income",
                  "occupation" = "Occupation",
                  "PA" = "Physical\n Activity"),

#identify exposure and outcome     
                  exposure = "NV",
                  outcome = "CVD")

#plot DAG using the ggdag function
ggdag(NV_CVD_dag, text = FALSE, use_labels = "label",edge_type="link_arc")

Identify Open Paths

Use the “ggdag_paths” function to identify all open paths (directed or non-directed). Consistent with our in-class example, we see six open paths. The first function might not be ideal, given that we have so many open paths. The “gg_paths_fan” (second example) is another option where you are able to overlay open paths.

ggdag_paths(NV_CVD_dag, text = FALSE, use_labels = "label", shadow = TRUE, 

#directed=false indicates all paths, regardless of direction            
            directed=FALSE,  

#changing node/text sizes for aesthetics             
            node_size = 8,   text_size = 2, stylized = TRUE)

#using "ggdag_paths_fan to overlap all paths
ggdag_paths_fan(NV_CVD_dag, text = FALSE, use_labels = "label", shadow = TRUE, 
            directed=FALSE,  
            node_size = 8,   text_size = 3, stylized = TRUE)

Identify Directed Paths

Use the “ggdag_paths” function to identify directed paths only.

ggdag_paths(NV_CVD_dag, text = FALSE, use_labels = "label", shadow = TRUE, 

#directed=true indicates directed paths            
            directed=TRUE,  
            node_size = 10,   text_size = 2.8, stylized = TRUE)

Identify Minimally Sufficient Set

Use the “ggdag_adjustment_set” function to identify the minimally sufficient set. As previously identified, {income} is the minimally sufficient set.

ggdag_adjustment_set(NV_CVD_dag, text = FALSE, use_labels = "label", 
                     shadow = TRUE, stylized = TRUE, node_size = 10,   text_size = 2.8)

Identify d-separated & d-connected Paths

Use the “ggdag_drelationship” function to identify d-separated and d-connected paths after conditioning on the minimally sufficient set (income).

ggdag_drelationship(NV_CVD_dag, controlling_for  = "income", text = FALSE, use_labels = "label", 
                    stylized = TRUE, node_size = 10,   text_size = 2.8, edge_type="link_arc")

The Basics of Building DAGs in R

Leah Moubadder