The central rule: a pending case is evidence. It tells us a decision had not happened by its last observation date, so excluding it would systematically make waits look shorter.
Competing risks, not a trend line
The forecast uses the non-parametric Aalen–Johansen estimator. Clearance and rejection are modeled as separate, competing outcomes; cases still pending on the data date are right-censored. No linear extrapolation is used.
Conditioning on time already waited
Probabilities compare only with cases that remained unresolved for at least as long as the current case.
Cohort selection
The model tries the requested detail within a two-year lookback and falls back to a broader cohort when statistical support is too small.
Decision window
Percentile dates describe time to either reported terminal decision. Unsupported percentiles are never invented.
Limitations
- Records are voluntarily reported and are not representative of every applicant.
- Changing policy, workload, and reporting behavior can make older cases less comparable.
- Fields may be incomplete or incorrect, and an individual case can differ for reasons not captured here.
- The result is not legal advice and cannot predict an official decision.
Source and contact
Most records originate from the community site linked here. Comments may be sent to the address shown below. checkee.info · [email protected]