torchcell.datamodels.schema.derive_se

derive_se(uncertainty: float | None, uncertainty_type: UncertaintyType | None, n_samples: int | None) → float | None

Derive the standard error of the mean from a reported uncertainty + its kind.

standard_error/bootstrap_se -> as-is (already an SE); sample_sd -> sd/sqrt(n); variance -> sqrt(var/n); ci95 -> half-width/1.96. n_samples is required for the kinds that divide. Returns None when nothing is reported.