Statistical assumptions
- Right censoring is assumed non-informative conditional on observed covariates.
- Row-level records are treated as independent observations.
- Parametric baseline assumes lognormal latent completion times.
- Tail-mixture update assumes external anchor fidelity at days 300 and 629.
Data quality assumptions
- Canonical row fields are publicly shareable and intentionally public.
- Missing-value sentinels (
?, FUTURE, N/A, #VALUE!) are interpreted as unknown.
- Duration columns are considered derived mirrors and validated against date arithmetic.
Interpretation limits
- This is a crowdsourced observational sample, not a complete population frame.
- Outputs are descriptive/probabilistic and not legal advice.
- Small sample size implies high variance, especially in tail behavior.