Identifiability

Identifiability asks whether model parameters can be uniquely or reliably determined from the available observations.

Structural identifiability

This is a property of the mathematical model and observation scheme under ideal, noise-free data. A structurally unidentifiable model can produce exactly the same observations from different parameter values.

Practical identifiability

A structurally identifiable parameter may still be estimated very imprecisely because real data are sparse, noisy or insufficiently informative.

Why it matters

An optimisation algorithm can return a numerical parameter value even when the data do not genuinely determine it.