torchcell.datamodels.schema.UncertaintyType

class UncertaintyType(*values)[source]

Bases: StrEnum

What a reported uncertainty number IS, so it converts to an SE correctly.

Rigor comes from naming the statistic: a bootstrap SD of an estimator is already an SE (never divide it by sqrt(n) again), whereas a sample SD of observations must be divided. There is deliberately NO unknown – strict labelling: if the kind is unknown we do not ingest the value.