Glossary#

Key terms used throughout TanaT documentation and API.

For an overview of how these concepts relate, see Core Concepts.

General terms#

Criterion#: A filtering rule applied to a data container (pool, sequence, trajectory) based on temporal, pattern, or feature conditions.
Entity#: The atomic unit of temporal data. An entity represents a single observation for one individual at a given point or period in time, and carries one or more entity features.
Entity feature#: A descriptive attribute of an entity, such as a visit type, a diagnosis code, or a lab result value. Features can be categorical or numerical.
Individual#: A unique subject of observation (patient, user, customer…). When an individual owns multiple sequences of different types, they together form a trajectory.
Metadata#: Descriptive information attached to a sequence or pool that characterises its temporal structure (granularity, timezone…) and the types of its entity features. TanaT infers metadata automatically and allows explicit overrides.
Pool#: A collection of sequences or trajectories across multiple individuals. Pools are the primary structure for batch analysis (distance matrices, clustering…).
Sequence#: An ordered collection of entities of the same type belonging to one individual, optionally enriched with static features. All entities share the same feature structure and temporal type (event, interval, or state).
Static feature#: A time-invariant attribute of an individual (e.g. birth date, gender, cohort). Can be attached to a sequence as well as to a trajectory.
Trajectory#: A collection of multiple sequences of different types for the same individual, optionally enriched with static features. Trajectories give a multidimensional view of an individual’s temporal evolution.
Zeroing#: The process of aligning sequences or trajectories to a common reference date (T0 / index date), transforming absolute timestamps into relative ones to enable meaningful cross-individual comparison. See tanat.zeroing.

Sequence types#

Event sequence#: A sequence of punctual events; each entity occurs at a single timestamp with no duration.
Interval sequence#: A sequence of duration-based entities with explicit start and end dates. Intervals can overlap.
State sequence#: A sequence of contiguous, non-overlapping states. Together the entities cover the full observation period without gaps.