Research pilot applications are open. Apply for pilot →
Concepts & guides

Longitudinal datasets

Configure repeated observations so Ounias can evaluate state changes, transitions, timing, and trajectories.

Required structure

Longitudinal analysis requires observations belonging to identifiable entities and a meaningful ordering.

  • Entity identifier
  • At least two observations for an eligible entity
  • Observation index or observation datetime

Observation ordering

Ounias needs to know the order in which observations occurred.

An observation index can provide explicit ordering. A datetime can provide both ordering and temporal information.

Where datetime information is available, additional timing and dwell analyses can be calculated.

Relationship to labeling

Longitudinal analysis operates on states produced by a compatible parent analysis.

The meaning of a trajectory therefore depends partly on the labeling model or representation from which its states were produced.

What Ounias derives

  • Label sequences
  • Transitions
  • Jump counts
  • Start and terminal state profiles
  • Temporal profiles
  • Dwell information where available
  • Movement and trajectory information