Comparisons with Other Packages

In this vignette, we compare caugi to some of the most widely used graph packages in R, Python, and Java.

Overview

The following table summarizes the packages included in this comparison, their general focus, and the languages they are packaged for.

An overview of the packages included in this comparison

Package	Type	Language	URL
igraph	General-purpose	R, Python, C	https://igraph.org
graph	General-purpose	R	https://github.com/Bioconductor/graph
gRbase	Graphical models	R	https://CRAN.R-project.org/package=gRbase
pcalg	Causal graphs	R	https://pcalg.r-forge.r-project.org
dagitty	Causal graphs	R, Web	https://www.dagitty.net
bnlearn	Bayesian networks	R	https://www.bnlearn.com
ggm	Graphical Markov models	R	https://CRAN.R-project.org/package=ggm
MixedGraphs	Causal graphs	R	https://github.com/rje42/MixedGraphs
NetworkX	General-purpose	Python	https://networkx.org
pgmpy	Probabilitistc graphical models	Python	https://github.com/pgmpy/pgmpy
Tetrad	Causal graphs	Java, CLI, R1, Python	http://www.phil.cmu.edu/tetrad

Scope

The comparison focuses on graph representation and analysis: which graph classes each package can represent, which structural and causal-graph algorithms it implements, and how it interoperates with other tooling.

Two adjacent areas are intentionally out of scope:

• Causal discovery (PC, FCI, GES, GFCI, LiNGAM, etc.). caugi does not implement discovery algorithms.
• Statistical inference and parameter learning (CPT estimation, likelihood, parameter fitting): also out of scope for caugi.

Graph Types Supported

Supported graph types for the packages in this comparison. ● indicates a dedicated class with type-level invariants; ◐ indicates representability without a dedicated class or invariant enforcement; ○ indicates unsupported graph types.

Package	DAG	CPDAG	MPDAG	MAG	PAG	ADMG	SWIG	UG	Mixed/general
caugi	●	●2	●	◐3	○	●	○	●	●
igraph	◐4	○5	○	○	○	○	○	◐6	◐7
graph	◐8	○	○	○	○	○	○	◐9	◐10
gRbase	◐11	○	○	○	○	○	○	◐12	○
pcalg	◐13	◐14	◐15	◐16	◐17	○	○	○	○
dagitty	●	◐18	○	●	●	○	○	○	○
bnlearn	●	◐19	○	○	○	○	○	◐20	◐21
ggm	◐22	◐23	○	◐24	○	◐25	○	◐26	◐27
MixedGraphs	◐28	○	○	○	○	◐29	○	●	●
NetworkX	◐30	○	○	○	○	○	○	◐31	◐32
pgmpy	●	◐33	○34	●	○	●	○	●	○35
Tetrad	●	◐36	◐37	◐38	●39	◐40	○	◐41	◐42

Graph Queries and Structural Operations

Overview of available graph queries and structural operations in the packages in the comparison.

Package	Parents/children	Ancestors/descendants	d-sep	m-sep	Paths	Acyclicity	Markov blanket	Moralization	Skeleton	v-structures	MEC enumeration
caugi	●	●	●	●	○	●	●	●	●	○	○
igraph	◐43	◐44	○	○	●	●	○	○	○	○	○
graph	◐45	◐46	○	○	◐47	○	○	○	○	○	○
gRbase	◐48	◐49	○	○	○	●	○	●	○	○	○
pcalg	◐50	◐51	◐52	◐53	○	◐54	○	○	◐55	◐56	●
dagitty	●	●	●	○	●	●	●	●	○	◐57	●
bnlearn	●	●	◐58	○	●	●	●	●	●	●	●
ggm	◐59	◐60	●	●	◐61	●	○	○	○	○	◐62
MixedGraphs	●	●	○	●	○	◐63	◐64	●	●	○	○
NetworkX	◐65	●	◐66	○	●	●	○	●	◐67	●	○
pgmpy	●	◐68	●	●	○	●	●	●	○69	●	○
Tetrad	●	●	◐70	●	●	◐71	●	○	◐72	◐73	●

Causal-Inference Algorithms

Overview of available causal-inference algorithms in the packages in the comparison.

Package	Back-door adj.	Generalized adj.	Optimal adj.	ID algorithm	Do-calculus	Counterfactuals	Interventions/mutilation
caugi	●	●	◐74	○	○	○	○
igraph	○	○	○	○	○	○	○
graph	○	○	○	○	○	○	○
gRbase	○	○	○	○	○	○	○
pcalg	◐75	●	●	○	○	○	◐76
dagitty	◐77	○	○	○	○	○	◐78
bnlearn	○	○	○	○	○	●	●
ggm	○	○	○	○	○	○	○
MixedGraphs	○	○	○	○	○	○	◐79
NetworkX	○	○	○	○	○	○	○
pgmpy	●	●	○	○	○	○	●
Tetrad	◐80	●	◐81	○	○	○	○

I/O and Interoperability

Comparison of supported graph I/O formats and interoperability features.

Package	DOT	Mermaid	GraphML	JSON	Coerce to/from other graph classes
caugi	◐82	●	●	●83	●
igraph	◐84	○	●	○	◐85
graph	◐86	○	○	○	◐87
gRbase	○	○	○	○	◐88
pcalg	○	○	○	○	◐89
dagitty	○	○	○	○	◐90
bnlearn	◐91	○	○	○	●
ggm	○	○	○	○	◐92
MixedGraphs	○	○	○	○	●
NetworkX	◐93	○	●	●	◐94
pgmpy	◐95	○	○	○	◐96
Tetrad	◐97	○	○	●	○

Implementation and Ergonomics

Comparison of implementation details and ergonomic features of the packages in the comparison.

Package	Backend	Sparse storage	Built-in plotting	Layout algorithms	Pipeable/fluent API	Lazy mutation
caugi	R + Rust	●	●	●	●	●
igraph	C	◐98	●	●	○	○
graph	R + C	●99	○100	○	○	○
gRbase	R + C++	●	◐101	○	○	○
pcalg	R + C++	○	◐102	○	○	○
dagitty	R + JS	○	●	◐103	○	○
bnlearn	R + C	◐104	◐105	◐106	○	○
ggm	R	○	●	◐107	○	○
MixedGraphs	R + C++	◐108	◐109	○	◐110	○
NetworkX	Python	◐111	◐112	●	○	○
pgmpy	Python (NetworkX)	○	◐113	◐114	○	○
Tetrad	Java	●	◐115	●	○	○

Contributing

If you find any errors in the comparison above or would like to add another package for comparisons, please file an issue or submit a pull request with the relevant information.