45. SingleGraphAlg (Imported Graph Wrapper)

SingleGraphAlg is not a causal discovery algorithm in the usual sense.
Instead, it is a thin wrapper that lets you treat a fixed, user-supplied graph as if it were the output of a Tetrad algorithm.

This is useful when:

  • You have already run a causal discovery method outside Tetrad (e.g., in R, Python, or another tool) and want to:

    • Compare that graph to Tetrad’s algorithms in the Simulation / Algorithm Comparison workflow, or

    • Use it as a “baseline” or reference graph in performance tables.

  • You have constructed a graph by hand (e.g., from domain knowledge or a published model) and want it to appear in the same tables/plots as algorithm outputs.

  • You want to include a “gold standard” graph in evaluations, but that gold standard lives in a file.

45.1. What it does

  • You load a graph from a file in one of Tetrad’s supported graph formats.

  • SingleGraphAlg simply returns that graph unchanged as its “search result.”

  • From the rest of Tetrad’s perspective, this looks just like the output of any other algorithm:

    • It can be included in graph comparison tables (adjacency/arrowhead/tail metrics, etc.).

    • It can be visualized, saved, and used in downstream workflows that expect an algorithm-produced graph.

Crucially:

  • No data are used. SingleGraphAlg completely ignores the input dataset.

  • No edges are added, removed, or reoriented. The imported graph is taken as-is.

45.2. Typical workflow

  1. Create or learn a graph outside Tetrad, then export it to one of Tetrad’s supported text graph formats (see Graph Formats).

  2. In Tetrad, select SingleGraphAlg as one of the algorithms in your comparison.

  3. Use the algorithm’s configuration to point it at your graph file.

  4. Run the comparison as usual:

    • SingleGraphAlg will contribute exactly the imported graph.

    • Other algorithms will learn graphs from the data.

    • All graphs appear side by side in the evaluation tables and plots.

45.3. When to use (and when not to)

Use SingleGraphAlg when:

  • You want an external or hand-built graph to participate in Tetrad’s evaluation and visualization pipeline.

  • You are doing benchmarking and want to include:

    • A model from another package (e.g., dagitty, pcalg, bnlearn, custom code), or

    • A known “true” graph stored on disk.

Do not use SingleGraphAlg if:

  • You actually want Tetrad to learn a graph from data—choose PC, FGES, GFCI, FCIT, BOSS, etc. instead.

  • You expect it to modify or repair the imported graph. It will not; it trusts the file as given.