{"id":"arxiv/2611.11874","url":"https://arxiv.org/abs/2611.11874","tags":["math.ST","math.PR","stat.ME"],"score":0,"title":"An iterated-logarithm price of anytime validity for log-optimal e-processes","domain":"math.ST","source":"arxiv","authors":["Adrien Corenflos"],"reviews":0,"subject":"math.ST","summary":"Log-optimal e-variables at fixed horizons, also called numéraires, have been conjectured to ``almost'' form an e-process for exponential-family null hypotheses. A related open question is whether an e-process can attain the optimal e-power at every horizon up to a bounded loss. Given a final horizon, we study the minimax regret of e-processes relative to the numéraires at all times up to it. For the Gaussian location null and the composite alternative of Gaussian distributions with a smaller variance and an unknown location, we show that the minimax regret in the worst case over the location…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-11-09","reactions":{}}