Tests & Scores

Causal discovery in Tetrad relies on conditional-independence tests (for constraint-based methods) and scores (for score-based or hybrid methods). Choosing the right test or score is essential to obtaining accurate and reliable graphs.

This section provides:

  • A practical guide to choosing an appropriate test/score for your data type
    β†’ Choosing Tests & Scores

  • A catalog-style reference summarizing all available tests and scores in Tetrad
    β†’ Tests & Scores Catalog

Use these pages together when selecting modeling assumptions for PC, FGES, BOSS, GFCI, and related algorithms.

A few tests and scores are marked experimental, are hidden from the chooser by default, and are offered without a documentation commitment. See Experimental Methods for what the marking means and for the current list.