Anytime validity is free: inducing sequential tests
Journal of the Royal Statistical Society: Series B (Statistical Methodology) · 2026 · Volume 88 · Issue 4
Abstract
Anytime valid sequential tests permit us to stop testing based on the current data, without invalidating the inference. Given a maximum number of observationsN, one may believe this must come at the cost of power when compared to a conventional test that waits until allNobservations have arrived. Our first contribution is to show that this is false: for any valid test based onNobservations, we show how to construct an anytime valid sequential test that matches it afterNobservations. Our second contribution is that we may use the outcome of a[0,1]-valued test as a conditional significance level in subsequent testing, leading to an overall procedure that is valid at the original significance level. This shows that anytime validity and optional continuation are readily available in traditional testing, without requiring explicit use ofe-values. We illustrate this by deriving the anytime valid sequentializedz-test andt-test, which at timeNcoincide with the traditionalz-test andt-test. Finally, we characterize the Sequential Probability Ratio Test by invariance under test induction, and show that it is induced by the Neyman–Pearson test for a tiny level and hugeNunder an i.i.d. assumption.
Keywords
anytime validity, e-values, optional continuation, randomized testing, sequentialt-test, sequential testing
Enable JavaScript for citation tools and additional paper interactions.