Oncesignificant

Oncesignificant is a term used in statistics and data science to describe a result that has been judged statistically significant in a single analysis or stage but does not remain significant under replication or subsequent analyses. It signals conditional or ephemeral significance rather than a robust, replicated effect.

In sequential studies, an oncesignificant finding may cross a pre-specified significance threshold at an interim analysis but lose significance when additional data are incorporated. In cross-validation or high-dimensional testing, a feature or outcome labeled oncesignificant may appear significant in one fold or comparison yet fail to replicate across folds or after adjusting for multiple comparisons.

Origin and usage: The term is a neologism that has appeared primarily in informal statistics discussions, preprints, and blog posts in the 2020s. It is not widely standardized in formal statistical texts, and its exact interpretation can vary by field or study design.

Interpretation and cautions: Labeling a result as oncesignificant emphasizes the need for replication and robust evidence. It is not a substitute for preregistration, proper multiple-testing correction, or replication studies. Critics warn that such a label may encourage premature conclusions if used without clear criteria.

See also: p-value, statistical significance, multiple comparisons, replication crisis, preregistration.