DAGassist in the R DAG Ecosystem

Introduction

This article explains the broader R DAG ecosystem and positions DAGassist within it. It also explains the dependencies necessary to maximize DAGassist’s functionality.

• dagitty for DAG specification and reasoning
• ggdag to visualize DAGs
• modelsummary to format regression tables
• broom to tidy models (coefs & goodness-of-fit)
• writexl to export multi‑sheet Excel workbooks
• rmarkdown to render shareable docs (HTML/Word/PDF) that embed DAGassist outputs

The R DAG Ecosystem at a Glance

The Differences Between dagitty, ggdag, and DAGassist

dagitty

• The core DAG toolkit for R. Creates the actual DAG object that ggdag and DAGassist use.
• Supports causal queries:
  • dagitty::adjustmentSets() scans a dagitty object to identify adjustment sets.
  • dagitty::isAcyclic() verifies acyclicity.
  • dagitty::findCycle() identifies cyclical variable relationships.

Making a DAG with dagitty:

#via dagitty
library(dagitty)

dag_model <- dagitty('dag {
    EffectiveN_electoral
    Gov_effectiveness_0_5
    LR_Polarization_perceived
    PID_noleaners [exposure]
    Presidentialsystem
    VoterAPI_Party [outcome]
    Year_0
    EffectiveN_electoral -> VoterAPI_Party
    Gov_effectiveness_0_5 -> VoterAPI_Party
    LR_Polarization_perceived -> VoterAPI_Party
    PID_noleaners -> LR_Polarization_perceived
    PID_noleaners -> VoterAPI_Party
    Presidentialsystem -> EffectiveN_electoral
    Presidentialsystem -> VoterAPI_Party
    Year_0 -> PID_noleaners
    Year_0 -> VoterAPI_Party
}')  

ggdag

• A visualization layer for DAGs.
• Also provides a more intuitive way to make dagitty objects.

Making an identical DAG with ggdag:

#via ggdag
library(ggdag)

dag_model <- dagify(
  VoterAPI_Party  ~ PID_noleaners + LR_Polarization_perceived +
    Gov_effectiveness_0_5 + EffectiveN_electoral + Presidentialsystem,
    
  LR_Polarization_perceived ~  PID_noleaners,
  
  EffectiveN_electoral ~ Presidentialsystem,
    
  exposure = "PID_noleaners",
  outcome = "VoterAPI_Party"
)

DAGassist

• The glue between your DAG models and regressions.
• Uses dagitty logic to classify the causal relationships embedded in a DAG.
• Identifies those relationships and their implications for your covariate selection in a concise, publication-grade report.

It uses dagitty logic to derive minimal and canonical adjustment sets, according to these criteria:

Path / Node Type	Minimal	Canonical
Fork/Common–Cause Confounder (Z)	✓	✓
Chain/Mediator (M)	✗	✗
Collider (C)	✗	✗
Descendant of Mediator (N)	✗	✗
Descendant of Collider (Q)	✗	✗
Descendant of Outcome (I)	✗	✗
M-Bias	✗	✗
Butterfly Bias	✗	✗
Neutral Control on Treatment (E → X)	✗	✓
Neutral Control on Outcome (F → Y)	✗	✓
Descendant of Confounder off Backdoor Path (W)	✗	✗
Descendant of Confounder on Backdoor Path (V)	Z or V	Z and V

Note: ✓ = adjust; ✗ = do not adjust. There may be multiple minimal sets; the canonical set is unique.