Positivity, overlap & extrapolation
6 open problems
When weights explode or a model extrapolates, the nominal target estimand may no longer be identified. Effective sample size does not show this.
| ID | Problem | Priority | Verdict | Maturity | prank |
|---|---|---|---|---|---|
| OVL-01 | Weak overlap leaves target regions the source data cannot support | Very high | Partially addressed | Established | 1 |
| OVL-02 | ESS is an insufficient overlap diagnostic and no battery is agreed | Very high | Partially addressed | Promising | 1 |
| OVL-04 | Weight trimming silently changes the estimand | Very high | Overstated | Emerging | 1 |
| OVL-03 | No support diagnostic aggregates over a PAIC target distribution | High | Partially addressed | Emerging | 2 |
| OVL-05 | Optimal transport for distributional balancing in PAIC | High | Confirmed open | Promising | 2 |
| OVL-06 | No standardized sample-size planning guidance or software for PAIC | High | Confirmed open | Emerging | 2 |
No matching items