How DAGMakie compares

DAGMakie.jl draws causal DAGs, mixed graphs, and related diagrams with Makie. It is visualisation only: identification and estimation stay in CausalDynamics.jl / CausalInference.jl and CausalTargeted.jl.

Closest R tools are ggdag and dagitty plots. In Python, people typically glue networkx, matplotlib, graphviz, or CausalGraphicalModels.

Choose DAGMakie when you want layered causal themes, bidirected confounding, path highlighting, time-indexed layouts, or SWIG / DiD visual grammar in the same Makie stack as SciML and book figures.

Prefer ggdag / dagitty when your analysis is already tidyverse-centric, you need the dagitty web GUI, or a one-off ggplot is enough.

Stack overview: ECOSYSTEM_COMPARISON.md.

flowchart LR
  subgraph juliaStack [Julia_typed_pipeline]
    G[Graph_CDM] --> ID[identify_certificate]
    ID --> SIM[simulate_counterfactual]
    ID --> EST[LMTP_mediation]
    ID --> PLOT[DAGMakie]
    SIM --> EST
  end

Legend

Mark	Meaning
`Yes`	First-class, documented
`Partial`	Possible with glue or a limited API
`—`	Not in that package’s usual scope
`Unique`	Strong differentiator here

Versus R and Python (DAG visualisation)

Capability	DAGMakie	R	Python
Layered causal DAG plot	Yes	Yes (ggdag, dagitty)	Partial (networkx / graphviz / CGMs)
Bidirected confounding arcs	Yes	Yes (ggdag / dagitty)	Partial
Path / adjustment / d-sep highlight	Yes (± CausalInference)	Yes (ggdag + dagitty)	Partial (DoWhy / CGMs)
Smart / dagitty-like ancestor colouring	Yes (`smart=true`)	Yes (dagitty)	Partial
Display-only `do(·)` comparison figures	Yes	Partial (custom)	Partial (custom)
Time-indexed / unrolled layouts	Yes	Partial (custom ggplot)	Partial (custom)
SWIG / DiD / IDAG visual grammar	Unique	Partial (custom)	Partial (custom)
Same Makie stack as SciML figures	Unique	—	—
Graph GUI / interactive editor	—	Yes (dagitty web)	Partial

Julia neighbours

Package	Role
GraphMakie.jl	General graph plotting backend
CausalDynamics.jl	Identification sets consumed by plot helpers (comparison)
CausalTargeted.jl	Estimation (not plotted here) (comparison)

What is distinctive here

Causal diagram defaults — steel-blue themes, node roles, bidirected geometry
Visual grammar — IDAG effect-measure nodes and DiD SWIG layouts (guide)
Pipeline fit — figures sit next to SciML and Documenter @example chunks without leaving Julia

What is deliberately absent

Identification algorithms, estimation, and a dagitty-style GUI. Pass sets into dagplot_*, or load CausalInference / CausalDynamics when the plot should compute highlighting. See Causal analysis.

The CDCS book uses these figures in narrative context.