Glossary

Term Meaning
STACK System for Teaching and Assessment using a Computer algebra Kernel, a Moodle question type that grades mathematics by evaluating a student's algebraic answer with a CAS (Maxima), rather than matching text.
PRT Potential Response Tree, the decision logic inside a STACK question that compares a student's answer against various tests and assigns a mark and feedback based on which branch of the tree the answer lands on.
Seed A random number Moodle assigns to an attempt at a randomized STACK question, determining which specific instantiated values (and therefore which "version") of the question a student receives.
Facility Index Moodle's own built-in measure of how easy a question was in practice, essentially its average score across attempts, on a 0 to 1 scale.
Pool A / Pool B This plugin's own internal terms for two different sets of response rows: Pool A is every attempt (participation); Pool B is one best attempt per student (performance). See How Everything Is Calculated § Two response pools.
Indicator In Moodle's Analytics API, a single normalized signal (bounded to [-1, 1]) that feeds into a prediction model, but is not itself the thing being predicted.
Target In Moodle's Analytics API, the actual outcome a model is trained to predict: in this plugin, either "will this student's grade fall below the pass threshold" (Model 1) or "does this question's pass rate suggest it needs review" (Model 2).
Diagnostic report A statistical calculation (seed bias, PRT branch coverage) deliberately kept outside the Analytics API's machine-learning pipeline, because it has no natural ground-truth label to predict against.
Tree edit distance (Zhang and Shasha) A standard algorithm for measuring how different two tree structures are, by counting the minimum number of node insertions, deletions, and renames needed to turn one into the other.
Cohen's d / η² bands Conventional thresholds (Jacob Cohen, 1988) for calling an effect size negligible, small, medium, or large, used here to interpret the seed-bias ANOVA's η² without needing an exact p-value.

Full context for every term above: How Everything Is Calculated.