## Problem 044 — rank:299

**Title:** Practical identifiability

**Added binding assumptions (LaTeX):**

```latex
Use an identifiable parameter representative or a metric on the symmetry quotient, rather than penalizing unobservable coordinate choices. Structural identifiability does not ensure uniform finite-sample precision near singular or weakly sensitive parameter values. State which likelihood regularity and separation assumptions yield attainable bounds; allow a positive minimax lower bound above $\varepsilon$.
```

**Direct edits:** 0. **Fresh external checks:** None; supplied sources retained as provenance.